def get_batch(self):
# slice roidb
cur_from = self.cur
cur_to = min(cur_from + self.batch_size, self.size)
roidb = [self.roidb[self.index[i]] for i in range(cur_from, cur_to)]
# decide multi device slice
work_load_list = self.work_load_list
ctx = self.ctx
if work_load_list is None:
work_load_list = [1] * len(ctx)
assert isinstance(work_load_list, list) and len(work_load_list) == len(ctx), \
"Invalid settings for work load. "
slices = _split_input_slice(self.batch_size, work_load_list)
# get testing data for multigpu
data_list = []
label_list = []
for islice in slices:
iroidb = [roidb[i] for i in range(islice.start, islice.stop)]
data, label = get_rpn_batch(iroidb, self.cfg)
data_list.append(data)
label_list.append(label)
# pad data first and then assign anchor (read label)
data_tensor = tensor_vstack([batch['data'] for batch in data_list])
for data, data_pad in zip(data_list, data_tensor):
data['data'] = data_pad[np.newaxis, :]
new_label_list = []
for data, label in zip(data_list, label_list):
# infer label shape
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape, label['gt_boxes'],
data['im_info'], self.cfg, self.class_num,
self.feat_stride, self.anchor_scales,
self.anchor_ratios, self.allowed_border,
self.multi_label)
new_label_list.append(label)
all_data = dict()
for key in self.data_name:
all_data[key] = tensor_vstack([batch[key] for batch in data_list])
all_label = dict()
for key in self.label_name:
pad = -1 if key == 'label' else 0
all_label[key] = tensor_vstack(
[batch[key] for batch in new_label_list], pad=pad)
self.data = [mx.nd.array(all_data[key]) for key in self.data_name]
self.label = [mx.nd.array(all_label[key]) for key in self.label_name]
def parfetch(self, iroidb):
# get testing data for multigpu
data, label = get_rpn_batch(iroidb, self.cfg)
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape_p3, _ = self.feat_sym_p3.infer_shape(**data_shape)
feat_shape_p3 = [int(i) for i in feat_shape_p3[0]]
_, feat_shape_p4, _ = self.feat_sym_p4.infer_shape(**data_shape)
feat_shape_p4 = [int(i) for i in feat_shape_p4[0]]
_, feat_shape_p5, _ = self.feat_sym_p5.infer_shape(**data_shape)
feat_shape_p5 = [int(i) for i in feat_shape_p5[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape_p3, feat_shape_p4, feat_shape_p5,
label['gt_boxes'], data['im_info'], self.cfg,
self.feat_stride_p3, self.anchor_scales_p3,
self.anchor_ratios_p3, self.feat_stride_p4,
self.anchor_scales_p4, self.anchor_ratios_p4,
self.feat_stride_p5, self.anchor_scales_p5,
self.anchor_ratios_p5, self.allowed_border)
return {'data': data, 'label': label}
def parfetch(self, roidb):
# get data for multi-gpu
data, label = get_rpn_batch(roidb, self.cfg)
data_shape = {k: v.shape for k, v in data.items()}
# not use im info for RPN training
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# add gt_masks to data for e2e
assert len(roidb) == 1
gt_masks = get_gt_masks(roidb[0]['cache_seg_inst'], data['im_info'][0,:2].astype('int'))
data['gt_masks'] = gt_masks
# assign anchor for label
label = assign_anchor(feat_shape, label['gt_boxes'], data['im_info'], self.cfg,
self.feat_stride, self.anchor_scales,
self.anchor_ratios, self.allowed_border)
# don't like the original name
new_label = {
'proposal_label': label['label'],
'proposal_bbox_target': label['bbox_target'],
'proposal_bbox_weight': label['bbox_weight'],
}
label = new_label
return {'data': data, 'label': label}
def parfetch(self, iroidb):
# get testing data for multigpu
#data, label = get_rpn_triple_batch(iroidb, self.cfg)
if config.TRAIN.END2END and config.TRAIN.E2E_NAME == "base": # resnet_rfcn
data, label = get_rpn_batch(iroidb, self.cfg)
elif config.TRAIN.END2END and config.TRAIN.E2E_NAME == "off":
data, label = get_rpn_pair_batch(iroidb, self.cfg)
elif config.TRAIN.END2END:
data, label = get_rpn_triple_batch(iroidb, self.cfg)
else:
data, label = get_rpn_triple_batch(iroidb, self.cfg)
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape, label['gt_boxes'], data['im_info'],
self.cfg, self.feat_stride, self.anchor_scales,
self.anchor_ratios, self.allowed_border,
self.normalize_target, self.bbox_mean,
self.bbox_std)
return {'data': data, 'label': label}
def get_batch(self):
# slice roidb
cur_from = self.cur
cur_to = min(cur_from + self.batch_size, self.size)
roidb = [self.roidb[self.index[i]] for i in range(cur_from, cur_to)]
# decide multi device slice
work_load_list = self.work_load_list
ctx = self.ctx
if work_load_list is None:
work_load_list = [1] * len(ctx)
assert isinstance(work_load_list, list) and len(work_load_list) == len(ctx), \
"Invalid settings for work load. "
slices = _split_input_slice(self.batch_size, work_load_list)
# get testing data for multigpu
data_list = []
label_list = []
for islice in slices:
iroidb = [roidb[i] for i in range(islice.start, islice.stop)]
data, label = get_rpn_batch(iroidb, self.cfg)
data_list.append(data)
label_list.append(label)
# pad data first and then assign anchor (read label)
data_tensor = tensor_vstack([batch['data'] for batch in data_list])
for data, data_pad in zip(data_list, data_tensor):
data['data'] = data_pad[np.newaxis, :]
new_label_list = []
for data, label in zip(data_list, label_list):
# infer label shape
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape, label['gt_boxes'], data['im_info'], self.cfg,
self.feat_stride, self.anchor_scales,
self.anchor_ratios, self.allowed_border,
self.normalize_target, self.bbox_mean, self.bbox_std)
new_label_list.append(label)
all_data = dict()
for key in self.data_name:
all_data[key] = tensor_vstack([batch[key] for batch in data_list])
all_label = dict()
for key in self.label_name:
pad = -1 if key == 'label' else 0
all_label[key] = tensor_vstack([batch[key] for batch in new_label_list], pad=pad)
self.data = [mx.nd.array(all_data[key]) for key in self.data_name]
self.label = [mx.nd.array(all_label[key]) for key in self.label_name]
def par_assign_anchor_wrapper(cfg, iroidb, feat_sym, feat_strides, anchor_scales, anchor_ratios, allowed_border):
data, rpn_label = get_rpn_batch(iroidb, cfg)
data_shape = {k:v.shape for k,v in data.items()}
del data_shape['im_info']
data['gt_boxes'] = rpn_label['gt_boxes'][np.newaxis,:,:]
feat_shape = [y[1] for y in [x.infer_shape(**data_shape) for x in feat_sym]]
label = assign_pyramid_anchor(feat_shape, rpn_label['gt_boxes'],data['im_info'],cfg,
feat_strides, anchor_scales, anchor_ratios, allowed_border)
return {'data':data,'label':label}
def get_batch_parallel(self):
cur_from = self.cur
cur_to = min(cur_from + self.batch_size, self.size)
roidb = [self.roidb[self.index[i]] for i in range(cur_from, cur_to)]
work_load_list = self.work_load_list
ctx = self.ctx
if work_load_list is None:
work_load_list = [1] * len(ctx)
slices = _split_input_slice(self.batch_size, work_load_list)
max_data = {}
max_label = {}
data_lst = []
rpn_label_lst = []
for idx, islice in enumerate(slices):
iroidb = [roidb[i] for i in range(islice.start, islice.stop)]
data, rpn_label = get_rpn_batch(iroidb, self.cfg)
data['gt_boxes'] = rpn_label['gt_boxes'][np.newaxis,:,:]
data_shape = {k:list(v.shape) for k,v in data.items()}
if max_data == {} :
max_data = data_shape
else:
#max_data = {k:np.where(max_data[k]>v,max_data[k],v) for k,v in data_shape.items() }
for k,v in data_shape.items():
max_data[k] = np.where(np.array(max_data[k])>np.array(data_shape[k]),np.array(max_data[k]),np.array(data_shape[k]))
data_lst.append(data)
rpn_label_lst.append(rpn_label)
for k,v in max_data.items():
max_data[k][0] = self.batch_size
self.data = [mx.nd.zeros(tuple(max_data['data'])),mx.nd.zeros(tuple(max_data['im_info'])),mx.nd.full(tuple(max_data['gt_boxes']),-1)]
del max_data['im_info']
del max_data['gt_boxes']
max_data = {k:tuple(v) for k,v in max_data.items()}
all_label = {}
for idx, islice in enumerate(slices):
feat_shape = [y[1] for y in [x.infer_shape(**max_data) for x in self.feat_sym]]
d = data_lst[idx]
self.data[0][idx,:d['data'].shape[1],:d['data'].shape[2],:d['data'].shape[3]] = d['data'][0]
self.data[1][idx,:d['im_info'].shape[1]] = d['im_info'][0]
self.data[2][idx,:d['gt_boxes'].shape[1],:d['gt_boxes'].shape[2]] = d['gt_boxes'][0]
label = assign_pyramid_anchor(feat_shape, rpn_label_lst[idx]['gt_boxes'],data_lst[idx]['im_info'],self.cfg,
self.feat_strides, self.anchor_scales, self.anchor_ratios, self.allowed_border)
if all_label == {}:
all_label = label
else:
for k,v in label.items():
all_label[k] = np.vstack([all_label[k],v])
self.label = [mx.nd.array(v) for k,v in all_label.items()]
def par_assign_anchor_wrapper(cfg, iroidb, feat_sym, feat_strides, anchor_scales, anchor_ratios, allowed_border):
# get testing data for multigpu
data, rpn_label = get_rpn_batch(iroidb, cfg)
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
# add gt_boxes to data for e2e
data['gt_boxes'] = rpn_label['gt_boxes'][np.newaxis, :, :]
feat_shape = [y[1] for y in [x.infer_shape(**data_shape) for x in feat_sym]]
label = assign_pyramid_anchor(feat_shape, rpn_label['gt_boxes'], data['im_info'], cfg,
feat_strides, anchor_scales, anchor_ratios, allowed_border)
return {'data': data, 'label': label}
def parfetch(self, iroidb):
# get testing data for multigpu
data, label = get_rpn_batch(iroidb, self.cfg)
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape, label['gt_boxes'], data['im_info'], self.cfg,
self.feat_stride, self.anchor_scales,
self.anchor_ratios, self.allowed_border)
return {'data': data, 'label': label}
def parfetch(self, iroidb):
# get testing data for multigpu
data, label = get_rpn_batch(iroidb, self.cfg)
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape, label['gt_boxes'], data['im_info'], self.cfg,
self.feat_stride, self.anchor_scales,
self.anchor_ratios, self.allowed_border,
self.normalize_target, self.bbox_mean, self.bbox_std)
return {'data': data, 'label': label}
def parfetch(self, iroidb):
# get testing data for multigpu
data, label = get_rpn_batch(iroidb, self.cfg)
data_shape = {k: v.shape for k, v in data.items()}
del data_shape['im_info']
_, feat_shape, _ = self.feat_sym.infer_shape(**data_shape)
feat_shape = [int(i) for i in feat_shape[0]]
# add gt_boxes to data for e2e
data['gt_boxes'] = label['gt_boxes'][np.newaxis, :, :]
# assign anchor for label
label = assign_anchor(feat_shape=feat_shape,
gt_boxes=label['gt_boxes'],
im_info=data['im_info'],
cfg=self.cfg,
class_num=self.class_num,
feat_stride=self.feat_stride,
scales=self.anchor_scales,
ratios=self.anchor_ratios,
allowed_border=self.allowed_border,
multi_label=self.multi_label)
return {'data': data, 'label': label}
