def load_model(sym):
if system_dict["gpu"]:
ctx = mx.gpu(int(system_dict["gpu"]))
else:
ctx = mx.cpu(0)
# load params
arg_params, aux_params = load_param(system_dict["params"], ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, system_dict["img_long_side"],
system_dict["img_long_side"])),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
return mod
def convert_net(sym, args):
# setup context
ctx = mx.cpu(0)
# weight_map = get_weight_map(args.step_old, args.is_bin_old,
# args.step_new, args.is_bin_new)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# forward
mod.save_checkpoint(args.save_prefix, epoch=0)
def load_model(sym):
'''
User function: Loads the trained model weights
Args:
sym (mxnet model): Mxnet model returned from set_network() function
Returns:
mxnet model: Model with trained weights
'''
if system_dict["gpu"]:
ctx = mx.gpu(int(system_dict["gpu"]))
else:
ctx = mx.cpu(0)
# load params
arg_params, aux_params = load_param(system_dict["params"], ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, system_dict["img_long_side"],
system_dict["img_long_side"])),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
return mod
def test_net(sym, imdb, args):
# print config
logger.info('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
ctx = mx.gpu(args.gpu)
# load testing data
test_data = TestLoader(imdb.roidb, batch_size=1, short=args.img_short_side, max_size=args.img_long_side,
mean=args.img_pixel_means, std=args.img_pixel_stds)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)), ('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# all detections are collected into:
# all_boxes[cls][image] = N x 5 array of detections in
# (x1, y1, x2, y2, score)
all_boxes = [[[] for _ in range(imdb.num_images)]
for _ in range(imdb.num_classes)]
# start detection
with tqdm(total=imdb.num_images) as pbar:
for i, data_batch in enumerate(test_data):
# forward
im_info = data_batch.data[1][0]
mod.forward(data_batch)
rois, scores, bbox_deltas = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
det = im_detect(rois, scores, bbox_deltas, im_info,
bbox_stds=args.rcnn_bbox_stds, nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
for j in range(1, imdb.num_classes):
indexes = np.where(det[:, 0] == j)[0]
all_boxes[j][i] = np.concatenate((det[:, -4:], det[:, [1]]), axis=-1)[indexes, :]
pbar.update(data_batch.data[0].shape[0])
# evaluate model
imdb.evaluate_detections(all_boxes)
def demo_net(sym, class_names, args):
# print config
print('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
if args.gpu:
ctx = mx.gpu(int(args.gpu))
else:
ctx = mx.cpu(0)
# load single test
im_tensor, im_info, im_orig = load_test(args.image, short=args.img_short_side, max_size=args.img_long_side,
mean=args.img_pixel_means, std=args.img_pixel_stds)
# generate data batch
data_batch = generate_batch(im_tensor, im_info)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)), ('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# forward
mod.forward(data_batch)
rois, scores, bbox_deltas = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
im_info = im_info[0]
# decode detection
det = im_detect(rois, scores, bbox_deltas, im_info,
bbox_stds=args.rcnn_bbox_stds, nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
# print out
for [cls, conf, x1, y1, x2, y2] in det:
if cls > 0 and conf > args.vis_thresh:
print(class_names[int(cls)], conf, [x1, y1, x2, y2])
# if vis
if args.vis:
vis_detection(im_orig, det, class_names, thresh=args.vis_thresh)
def train(sym, roidb):
'''
User function: Start training
Args:
sym (mxnet model): Mxnet model returned from set_network() function
roidb (dataloader): Dataloader returned from set_model() function
Returns:
None
'''
# print config
#logger.info('called with system_dict\n{}'.format(pprint.pformat(vars(system_dict))))
#print(system_dict)
# setup multi-gpu
if(len(system_dict["gpus"]) == 0):
ctx = [mx.cpu(0)];
else:
ctx = [mx.gpu(int(i)) for i in system_dict["gpus"]]
batch_size = system_dict["rcnn_batch_size"] * len(ctx)
# load training data
feat_sym = sym.get_internals()['rpn_cls_score_output']
ag = AnchorGenerator(feat_stride=system_dict["rpn_feat_stride"],
anchor_scales=system_dict["rpn_anchor_scales"], anchor_ratios=system_dict["rpn_anchor_ratios"])
asp = AnchorSampler(allowed_border=system_dict["rpn_allowed_border"], batch_rois=system_dict["rpn_batch_rois"],
fg_fraction=system_dict["rpn_fg_fraction"], fg_overlap=system_dict["rpn_fg_overlap"],
bg_overlap=system_dict["rpn_bg_overlap"])
train_data = AnchorLoader(roidb, batch_size, system_dict["img_short_side"], system_dict["img_long_side"],
system_dict["img_pixel_means"], system_dict["img_pixel_stds"], feat_sym, ag, asp, shuffle=True)
# produce shape max possible
_, out_shape, _ = feat_sym.infer_shape(data=(1, 3, system_dict["img_long_side"], system_dict["img_long_side"]))
feat_height, feat_width = out_shape[0][-2:]
rpn_num_anchors = len(system_dict["rpn_anchor_scales"]) * len(system_dict["rpn_anchor_ratios"])
data_names = ['data', 'im_info', 'gt_boxes']
label_names = ['label', 'bbox_target', 'bbox_weight']
data_shapes = [('data', (batch_size, 3, system_dict["img_long_side"], system_dict["img_long_side"])),
('im_info', (batch_size, 3)),
('gt_boxes', (batch_size, 100, 5))]
label_shapes = [('label', (batch_size, 1, rpn_num_anchors * feat_height, feat_width)),
('bbox_target', (batch_size, 4 * rpn_num_anchors, feat_height, feat_width)),
('bbox_weight', (batch_size, 4 * rpn_num_anchors, feat_height, feat_width))]
# print shapes
data_shape_dict, out_shape_dict = infer_data_shape(sym, data_shapes + label_shapes)
logger.info('max input shape\n%s' % pprint.pformat(data_shape_dict))
logger.info('max output shape\n%s' % pprint.pformat(out_shape_dict))
# load and initialize params
if system_dict["resume"]:
arg_params, aux_params = load_param(system_dict["resume"])
else:
arg_params, aux_params = load_param(system_dict["pretrained"])
arg_params, aux_params = initialize_frcnn(sym, data_shapes, arg_params, aux_params)
# check parameter shapes
check_shape(sym, data_shapes + label_shapes, arg_params, aux_params)
# check fixed params
fixed_param_names = get_fixed_params(sym, system_dict["net_fixed_params"])
logger.info('locking params\n%s' % pprint.pformat(fixed_param_names))
# metric
rpn_eval_metric = RPNAccMetric()
rpn_cls_metric = RPNLogLossMetric()
rpn_bbox_metric = RPNL1LossMetric()
eval_metric = RCNNAccMetric()
cls_metric = RCNNLogLossMetric()
bbox_metric = RCNNL1LossMetric()
eval_metrics = mx.metric.CompositeEvalMetric()
for child_metric in [rpn_eval_metric, rpn_cls_metric, rpn_bbox_metric, eval_metric, cls_metric, bbox_metric]:
eval_metrics.add(child_metric)
# callback
batch_end_callback = mx.callback.Speedometer(batch_size, frequent=system_dict["log_interval"], auto_reset=False)
epoch_end_callback = mx.callback.do_checkpoint(system_dict["save_prefix"])
# learning schedule
base_lr = system_dict["lr"]
lr_factor = 0.1
lr_epoch = [int(epoch) for epoch in system_dict["lr_decay_epoch"].split(',')]
lr_epoch_diff = [epoch - system_dict["start_epoch"] for epoch in lr_epoch if epoch > system_dict["start_epoch"]]
lr = base_lr * (lr_factor ** (len(lr_epoch) - len(lr_epoch_diff)))
lr_iters = [int(epoch * len(roidb) / batch_size) for epoch in lr_epoch_diff]
logger.info('lr %f lr_epoch_diff %s lr_iters %s' % (lr, lr_epoch_diff, lr_iters))
lr_scheduler = mx.lr_scheduler.MultiFactorScheduler(lr_iters, lr_factor)
# optimizer
optimizer_params = {'momentum': 0.9,
'wd': 0.0005,
'learning_rate': lr,
'lr_scheduler': lr_scheduler,
'rescale_grad': (1.0 / batch_size),
'clip_gradient': 5}
# train
mod = Module(sym, data_names=data_names, label_names=label_names,
logger=logger, context=ctx, work_load_list=None,
fixed_param_names=fixed_param_names)
mod.fit(train_data, eval_metric=eval_metrics, epoch_end_callback=epoch_end_callback,
batch_end_callback=batch_end_callback, kvstore='device',
optimizer='sgd', optimizer_params=optimizer_params,
arg_params=arg_params, aux_params=aux_params, begin_epoch=system_dict["start_epoch"], num_epoch=system_dict["epochs"])
def train_net(sym, roidb, args):
# print config
logger.info('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup multi-gpu
ctx = [mx.cpu()] if not args.gpus else [
mx.gpu(int(i)) for i in args.gpus.split(',')
]
batch_size = args.rcnn_batch_size * len(ctx)
# load training data
feat_sym = sym.get_internals()['rpn_cls_score_output']
ag = AnchorGenerator(feat_stride=args.rpn_feat_stride,
anchor_scales=args.rpn_anchor_scales,
anchor_ratios=args.rpn_anchor_ratios)
asp = AnchorSampler(allowed_border=args.rpn_allowed_border,
batch_rois=args.rpn_batch_rois,
fg_fraction=args.rpn_fg_fraction,
fg_overlap=args.rpn_fg_overlap,
bg_overlap=args.rpn_bg_overlap)
train_data = AnchorLoader(roidb,
batch_size,
args.img_short_side,
args.img_long_side,
args.img_pixel_means,
args.img_pixel_stds,
feat_sym,
ag,
asp,
shuffle=True)
# produce shape max possible
_, out_shape, _ = feat_sym.infer_shape(data=(1, 3, args.img_long_side,
args.img_long_side))
feat_height, feat_width = out_shape[0][-2:]
rpn_num_anchors = len(args.rpn_anchor_scales) * len(args.rpn_anchor_ratios)
data_names = ['data', 'im_info', 'gt_boxes']
label_names = ['label', 'bbox_target', 'bbox_weight']
data_shapes = [('data', (batch_size, 3,
args.img_long_side, args.img_long_side)),
('im_info', (batch_size, 3)),
('gt_boxes', (batch_size, 100, 5))]
label_shapes = [('label', (batch_size, 1, rpn_num_anchors * feat_height,
feat_width)),
('bbox_target', (batch_size, 4 * rpn_num_anchors,
feat_height, feat_width)),
('bbox_weight', (batch_size, 4 * rpn_num_anchors,
feat_height, feat_width))]
# print shapes
data_shape_dict, out_shape_dict = infer_data_shape(
sym, data_shapes + label_shapes)
logger.info('max input shape\n%s' % pprint.pformat(data_shape_dict))
logger.info('max output shape\n%s' % pprint.pformat(out_shape_dict))
# load and initialize params
if args.resume:
arg_params, aux_params = load_param(args.resume)
else:
arg_params, aux_params = load_param(args.pretrained)
arg_params, aux_params = initialize_frcnn(sym, data_shapes, arg_params,
aux_params)
# check parameter shapes
check_shape(sym, data_shapes + label_shapes, arg_params, aux_params)
# check fixed params
fixed_param_names = get_fixed_params(sym, args.net_fixed_params)
logger.info('locking params\n%s' % pprint.pformat(fixed_param_names))
# metric
rpn_eval_metric = RPNAccMetric()
rpn_cls_metric = RPNLogLossMetric()
rpn_bbox_metric = RPNL1LossMetric()
eval_metric = RCNNAccMetric()
cls_metric = RCNNLogLossMetric()
bbox_metric = RCNNL1LossMetric()
eval_metrics = mx.gluon.metric.CompositeEvalMetric()
for child_metric in [
rpn_eval_metric, rpn_cls_metric, rpn_bbox_metric, eval_metric,
cls_metric, bbox_metric
]:
eval_metrics.add(child_metric)
# callback
batch_end_callback = mx.callback.Speedometer(batch_size,
frequent=args.log_interval,
auto_reset=False)
epoch_end_callback = mx.callback.do_checkpoint(args.save_prefix)
# learning schedule
base_lr = args.lr
lr_factor = 0.1
lr_epoch = [int(epoch) for epoch in args.lr_decay_epoch.split(',')]
lr_epoch_diff = [
epoch - args.start_epoch for epoch in lr_epoch
if epoch > args.start_epoch
]
lr = base_lr * (lr_factor**(len(lr_epoch) - len(lr_epoch_diff)))
lr_iters = [
int(epoch * len(roidb) / batch_size) for epoch in lr_epoch_diff
]
logger.info('lr %f lr_epoch_diff %s lr_iters %s' %
(lr, lr_epoch_diff, lr_iters))
lr_scheduler = mx.lr_scheduler.MultiFactorScheduler(lr_iters, lr_factor)
# optimizer
optimizer_params = {
'momentum': 0.9,
'wd': 0.0005,
'learning_rate': lr,
'lr_scheduler': lr_scheduler,
'rescale_grad': (1.0 / batch_size),
'clip_gradient': 5
}
# train
mod = Module(sym,
data_names=data_names,
label_names=label_names,
logger=logger,
context=ctx,
work_load_list=None,
fixed_param_names=fixed_param_names)
mod.fit(train_data,
eval_metric=eval_metrics,
epoch_end_callback=epoch_end_callback,
batch_end_callback=batch_end_callback,
kvstore='device',
optimizer='sgd',
optimizer_params=optimizer_params,
arg_params=arg_params,
aux_params=aux_params,
begin_epoch=args.start_epoch,
num_epoch=args.epochs)
def demo_net(sym, class_names, args):
# print config
print('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
if args.gpu:
ctx = mx.gpu(int(args.gpu))
else:
ctx = mx.cpu(0)
# load single test
im_tensor, im_info, im_orig = load_test(args.image,
short=args.img_short_side,
max_size=args.img_long_side,
mean=args.img_pixel_means,
std=args.img_pixel_stds)
# generate data batch
data_batch = generate_batch(im_tensor, im_info)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# forward
mod.forward(data_batch)
rois, scores, bbox_deltas = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
im_info = im_info[0]
# decode detection
det = im_detect(rois,
scores,
bbox_deltas,
im_info,
bbox_stds=args.rcnn_bbox_stds,
nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh,
use_soft_nms=args.use_soft_nms,
soft_nms_thresh=args.soft_nms_thresh,
max_per_image=args.max_per_image)
# print out
for [cls, conf, x1, y1, x2, y2] in det:
if cls > 0 and conf > args.vis_thresh:
print(class_names[int(cls)], conf, [x1, y1, x2, y2])
# if vis
if args.vis:
vis_detection(im_orig, det, class_names, thresh=args.vis_thresh)
def train_net(sym, roidb, args):
# print config
logger.info('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup multi-gpu
ctx = [mx.gpu(int(i)) for i in args.gpus.split(',')]
batch_size = args.rcnn_batch_size * len(ctx)
# load training data
feat_sym = sym.get_internals()['rpn_cls_score_output']
ag = AnchorGenerator(feat_stride=args.rpn_feat_stride,
anchor_scales=args.rpn_anchor_scales, anchor_ratios=args.rpn_anchor_ratios)
asp = AnchorSampler(allowed_border=args.rpn_allowed_border, batch_rois=args.rpn_batch_rois,
fg_fraction=args.rpn_fg_fraction, fg_overlap=args.rpn_fg_overlap,
bg_overlap=args.rpn_bg_overlap)
train_data = AnchorLoader(roidb, batch_size, args.img_short_side, args.img_long_side,
args.img_pixel_means, args.img_pixel_stds, feat_sym, ag, asp, shuffle=True)
# produce shape max possible
_, out_shape, _ = feat_sym.infer_shape(data=(1, 3, args.img_long_side, args.img_long_side))
feat_height, feat_width = out_shape[0][-2:]
rpn_num_anchors = len(args.rpn_anchor_scales) * len(args.rpn_anchor_ratios)
data_names = ['data', 'im_info', 'gt_boxes']
label_names = ['label', 'bbox_target', 'bbox_weight']
data_shapes = [('data', (batch_size, 3, args.img_long_side, args.img_long_side)),
('im_info', (batch_size, 3)),
('gt_boxes', (batch_size, 100, 5))]
label_shapes = [('label', (batch_size, 1, rpn_num_anchors * feat_height, feat_width)),
('bbox_target', (batch_size, 4 * rpn_num_anchors, feat_height, feat_width)),
('bbox_weight', (batch_size, 4 * rpn_num_anchors, feat_height, feat_width))]
# print shapes
data_shape_dict, out_shape_dict = infer_data_shape(sym, data_shapes + label_shapes)
logger.info('max input shape\n%s' % pprint.pformat(data_shape_dict))
logger.info('max output shape\n%s' % pprint.pformat(out_shape_dict))
# load and initialize params
if args.resume:
arg_params, aux_params = load_param(args.resume)
else:
arg_params, aux_params = load_param(args.pretrained)
arg_params, aux_params = initialize_frcnn(sym, data_shapes, arg_params, aux_params)
# check parameter shapes
check_shape(sym, data_shapes + label_shapes, arg_params, aux_params)
# check fixed params
fixed_param_names = get_fixed_params(sym, args.net_fixed_params)
logger.info('locking params\n%s' % pprint.pformat(fixed_param_names))
# metric
rpn_eval_metric = RPNAccMetric()
rpn_cls_metric = RPNLogLossMetric()
rpn_bbox_metric = RPNL1LossMetric()
eval_metric = RCNNAccMetric()
cls_metric = RCNNLogLossMetric()
bbox_metric = RCNNL1LossMetric()
eval_metrics = mx.metric.CompositeEvalMetric()
for child_metric in [rpn_eval_metric, rpn_cls_metric, rpn_bbox_metric, eval_metric, cls_metric, bbox_metric]:
eval_metrics.add(child_metric)
# callback
batch_end_callback = mx.callback.Speedometer(batch_size, frequent=args.log_interval, auto_reset=False)
epoch_end_callback = mx.callback.do_checkpoint(args.save_prefix)
# learning schedule
base_lr = args.lr
lr_factor = 0.1
lr_epoch = [int(epoch) for epoch in args.lr_decay_epoch.split(',')]
lr_epoch_diff = [epoch - args.start_epoch for epoch in lr_epoch if epoch > args.start_epoch]
lr = base_lr * (lr_factor ** (len(lr_epoch) - len(lr_epoch_diff)))
lr_iters = [int(epoch * len(roidb) / batch_size) for epoch in lr_epoch_diff]
logger.info('lr %f lr_epoch_diff %s lr_iters %s' % (lr, lr_epoch_diff, lr_iters))
lr_scheduler = mx.lr_scheduler.MultiFactorScheduler(lr_iters, lr_factor)
# optimizer
optimizer_params = {'momentum': 0.9,
'wd': 0.0005,
'learning_rate': lr,
'lr_scheduler': lr_scheduler,
'rescale_grad': (1.0 / batch_size),
'clip_gradient': 5}
# train
mod = Module(sym, data_names=data_names, label_names=label_names,
logger=logger, context=ctx, work_load_list=None,
fixed_param_names=fixed_param_names)
mod.fit(train_data, eval_metric=eval_metrics, epoch_end_callback=epoch_end_callback,
batch_end_callback=batch_end_callback, kvstore='device',
optimizer='sgd', optimizer_params=optimizer_params,
arg_params=arg_params, aux_params=aux_params, begin_epoch=args.start_epoch, num_epoch=args.epochs)
def dummy_data(ctx, batch_size=1):
return [
mx.nd.random.uniform(shape=shape, ctx=ctx)
for shape in ([batch_size, 3, 600, 600], [batch_size])
]
data_names = ['data']
label_names = None
data_shapes = [('data', (1, 3, 1000, 600))]
label_shapes = None
data = mx.symbol.Variable(name="data")
GLUON_LAYER = VGGConvBlock(isBin=True, step=4)
GLUON_LAYER.hybridize()
conv_feat = GLUON_LAYER(data)
arg_params, aux_params = load_param(
"/home/skutukov/work/mxnet_fasterrcnn_binary/convert/temp-0000.params",
ctx=mx.cpu())
check_shape(conv_feat, data_shapes, arg_params, aux_params)
mod = Module(conv_feat, data_names, label_names, context=mx.cpu())
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
data1, _ = dummy_data(ctx=mx.cpu())
# mod.forward(data1)
mod.save_checkpoint('test_vgg', epoch=0)
def test_net(sym, imdb, args):
# print config
logger.info('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
ctx = mx.gpu(args.gpu)
# load testing data
test_data = TestLoader(imdb.roidb,
batch_size=1,
short=args.img_short_side,
max_size=args.img_long_side,
mean=args.img_pixel_means,
std=args.img_pixel_stds)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# all detections are collected into:
# all_boxes[cls][image] = N x 5 array of detections in
# (x1, y1, x2, y2, score)
results_list = []
all_boxes = [[[] for _ in range(imdb.num_images)]
for _ in range(imdb.num_classes)]
all_masks = [[[] for _ in range(imdb.num_images)]
for _ in range(imdb.num_classes)]
all_rois = [[[] for _ in range(imdb.num_images)]
for _ in range(imdb.num_classes)]
# start detection
with tqdm(total=imdb.num_images) as pbar:
for i, data_batch in enumerate(test_data):
# forward
im_info = data_batch.data[1][0]
mod.forward(data_batch)
rois, scores, bbox_deltas, mask_prob = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
det, masks, rois_out = im_detect(rois,
scores,
bbox_deltas,
mask_prob,
im_info,
bbox_stds=args.rcnn_bbox_stds,
nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
# print(det.shape, masks.shape)
for j in range(1, imdb.num_classes):
indexes = np.where(det[:, 0] == j)[0]
all_boxes[j][i] = np.concatenate((det[:, -4:], det[:, [1]]),
axis=-1)[indexes, :]
# print(type(masks), type(rois_out))
all_masks[j][i] = masks[indexes]
all_rois[j][i] = rois_out[indexes]
boxes_this_image = [[]] + [
all_boxes[cls_ind][i]
for cls_ind in range(1, imdb.num_classes)
]
masks_this_image = [[]] + [
all_masks[cls_ind][i]
for cls_ind in range(1, imdb.num_classes)
]
rois_this_image = [[]] + [
all_rois[cls_ind][i] for cls_ind in range(1, imdb.num_classes)
]
results_list.append({
'image': '{}.png'.format(i),
'im_info': im_info.asnumpy(),
'boxes': boxes_this_image,
'masks': masks_this_image,
'rois': rois_this_image
})
pbar.update(data_batch.data[0].shape[0])
# evaluate model
results_pack = {
'all_boxes': all_boxes,
'all_masks': all_masks,
'results_list': results_list
}
imdb.evaluate_mask(results_pack)
data_names = [
'pair_im_1', 'pair_im_2', 'loca_map_1', 'loca_map_2', 'r_map',
'headsize_1', 'headsize_2'
]
label_names = None
data_shapes = [('pair_im_1', (1, 3, 250, 100)),
('pair_im_2', (1, 3, 250, 100)),
('loca_map_1', (1, 3, 250, 375)),
('loca_map_2', (1, 3, 250, 375)), ('r_map', (1, 3, 250, 375)),
('headsize_1', (1, 2)), ('headsize_2', (1, 2))]
label_shapes = None
# loading symbol model
sym = get_resnet_test_symbol(resnet_units, resnet_filter_list)
arg_params, aux_params = load_param(params_sym, ctx=ctx)
check_shape(sym, data_shapes, arg_params, aux_params)
mod = mx.mod.Module(symbol=sym,
context=ctx,
data_names=data_names,
label_names=label_names)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# test
tp_all, fp_all = 0, 0
acc_all = 0.0
for index_roi in range(len(roidb)):
start = time.time()
roidb_batch = roidb[index_roi]
def demo_net(sym, class_names, args):
# print config
# setup context
if args.gpu:
ctx = mx.gpu(int(args.gpu))
else:
ctx = mx.cpu(0)
print('called with args\n{}'.format(pprint.pformat(vars(args))))
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
f = open(
"/home/skutukov/datasets/VOCdevkit/VOC2007/ImageSets/Main/test.txt",
"r")
for file in tqdm.tqdm(f.readlines()):
path = os.path.join(args.image, str(file).strip() + '.jpg')
path = '/home/skutukov/Pictures/demo.jpg'
# load single test
im_tensor, im_info, im_orig = load_test(path,
short=args.img_short_side,
max_size=args.img_long_side,
mean=args.img_pixel_means,
std=args.img_pixel_stds,
ctx=ctx)
# generate data batch
data_batch = generate_batch(im_tensor, im_info)
# forward
mod.forward(data_batch)
rois, scores, bbox_deltas = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
im_info = im_info[0]
# decode detection
det = im_detect(rois,
scores,
bbox_deltas,
im_info,
bbox_stds=args.rcnn_bbox_stds,
nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
# print out
for [cls, conf, x1, y1, x2, y2] in det:
if cls > 0 and conf > args.vis_thresh:
print(class_names[int(cls)], conf, [x1, y1, x2, y2])
# if vis
if args.vis:
vis_detection(im_orig,
det,
class_names,
thresh=args.vis_thresh,
file=file)
break
# load training data
train_data = Loader(roidb, roihead, batch_size, img_pixel_means,
img_pixel_stds)
data_shape_dict, out_shape_dict = infer_data_shape(
net, train_data.provide_data + train_data.provide_label)
# load params
if is_init:
arg_params, aux_params = load_param_resnet_full(pretrained, ctx)
arg_params, aux_params = initialize_resnet_leaky(net,
train_data.provide_data,
arg_params, aux_params)
print('initialize and train the new network')
if is_resume:
arg_params, aux_params = load_param(resume, ctx)
print('resume from trained-network')
# symbol data
data_names = [
'pair_im_1', 'pair_im_2', 'loca_map_1', 'loca_map_2', 'r_map',
'headsize_1', 'headsize_2'
]
label_names = ['label']
check_shape(net, train_data.provide_data + train_data.provide_label,
arg_params, aux_params)
fixed_param_names = get_fixed_params(net, net_fixed_params)
# callback
def demo_net(sym, class_names, args):
# print config
print('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
if args.gpu:
ctx = mx.gpu(int(args.gpu))
else:
ctx = mx.cpu(0)
# load single test
im_tensor, im_info, im_orig = load_test(args.image, short=args.img_short_side, max_size=args.img_long_side,
mean=args.img_pixel_means, std=args.img_pixel_stds)
# generate data batch
data_batch = generate_batch(im_tensor, im_info)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)), ('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# forward
start=time.time()
mod.forward(data_batch)
rois, scores, bbox_deltas = mod.get_outputs()
print("time=", time.time() - start)
#rois = rois.asnumpy()
rois = rois[:, 1:]
#print('rois=',rois)
scores = scores[0]
bbox_deltas = bbox_deltas[0]
#print("BBox_deltas.shape=",bbox_deltas.shape)
#print("BBOX_deltas=",bbox_deltas)
im_info = im_info[0]
# decode detection
det = im_detect(rois, scores, bbox_deltas, im_info,
bbox_stds=args.rcnn_bbox_stds, nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
# print out
for [cls, conf, x_c,y_c,w,h,theta] in det:
if cls > 0 and conf > args.vis_thresh:
print('class_name=',class_names[int(cls)], 'conf=',conf, [x_c
, y_c, w,h,theta])
if True:
draw_rotate_box_cv(det,class_names,0.95)
def test_net(sym, imdb, args, config):
logger.addHandler(
logging.FileHandler("{0}/{1}".format(args.prefix, 'test.log')))
# print config
logger.info('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
ctx = mx.gpu(args.gpu)
# load testing data
test_data = TestLoader(imdb.roidb,
batch_size=1,
short=args.img_short_side,
max_size=args.img_long_side,
mean=config.transform['img_pixel_means'],
std=config.transform['img_pixel_stds'])
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# all detections are collected into:
# all_boxes[cls][image] = N x 5 array of detections in
# (x1, y1, x2, y2, score)
all_boxes = [[[] for _ in range(imdb.num_images)]
for _ in range(imdb.num_classes)]
# start detection
with tqdm(total=imdb.num_images) as pbar:
for i, data_batch in enumerate(test_data):
# forward
im_info = data_batch.data[1][0]
mod.forward(data_batch)
rois, scores, bbox_deltas = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
det = im_detect(rois,
scores,
bbox_deltas,
im_info,
bbox_stds=args.rcnn_bbox_stds,
nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
for j in range(1, imdb.num_classes):
indexes = np.where(det[:, 0] == j)[0]
all_boxes[j][i] = np.concatenate((det[:, -4:], det[:, [1]]),
axis=-1)[indexes, :]
pbar.update(data_batch.data[0].shape[0])
# evaluate model
imdb.evaluate_detections(all_boxes)
def demo_net(sym, class_names, args):
# print config
print('called with args\n{}'.format(pprint.pformat(vars(args))))
# setup context
if args.gpu:
ctx = mx.gpu(int(args.gpu))
else:
ctx = mx.cpu(0)
# load single test
im_tensor, im_info, im_orig = load_test(args.image,
short=args.img_short_side,
max_size=args.img_long_side,
mean=args.img_pixel_means,
std=args.img_pixel_stds)
# generate data batch
data_batch = generate_batch(im_tensor, im_info)
# load params
arg_params, aux_params = load_param(args.params, ctx=ctx)
# produce shape max possible
data_names = ['data', 'im_info']
label_names = None
data_shapes = [('data', (1, 3, args.img_long_side, args.img_long_side)),
('im_info', (1, 3))]
label_shapes = None
# check shapes
check_shape(sym, data_shapes, arg_params, aux_params)
# create and bind module
mod = Module(sym, data_names, label_names, context=ctx)
mod.bind(data_shapes, label_shapes, for_training=False)
mod.init_params(arg_params=arg_params, aux_params=aux_params)
# forward
mod.forward(data_batch)
rois, scores, bbox_deltas, mask_prob = mod.get_outputs()
rois = rois[:, 1:]
scores = scores[0]
bbox_deltas = bbox_deltas[0]
im_info = im_info[0]
# decode detection
det, masks = im_detect(rois,
scores,
bbox_deltas,
mask_prob,
im_info,
bbox_stds=args.rcnn_bbox_stds,
nms_thresh=args.rcnn_nms_thresh,
conf_thresh=args.rcnn_conf_thresh)
im = cv2.imread(args.image)
print(im.shape)
print(im_info)
# print out
for index, [cls, conf, x1, y1, x2, y2] in enumerate(det):
print(masks[index].max())
if cls > 0 and conf > args.vis_thresh:
print(class_names[int(cls)], conf, [x1, y1, x2, y2])
print((int(x1), int(y1)), (int(x2), int(y2)))
cv2.rectangle(im, (int(x1), int(y1)), (int(x2), int(y2)),
(255, 0, 0), 10)
cv2.imwrite("mask{}.png".format(index),
np.uint8(masks[index] * 255))
cv2.imwrite('demo.png', im)
# if vis
if args.vis:
vis_detection(im_orig, det, class_names, thresh=args.vis_thresh)
