def _get_kwargs(self):
# get kwargs of build_detection_train_loader
cfg = model_zoo.get_config("common/data/coco.py").dataloader.train
cfg.dataset.names = "coco_2017_val_100"
cfg.pop("_target_")
kwargs = {k: instantiate(v) for k, v in cfg.items()}
return kwargs
def get_model_no_weights(config_path):
"""
Like model_zoo.get, but do not load any weights (even pretrained)
"""
cfg = model_zoo.get_config(config_path)
if not torch.cuda.is_available():
cfg.MODEL.DEVICE = "cpu"
return build_model(cfg)
def testOmegaConf(self):
cfg = model_zoo.get_config(
"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml")
cfg = OmegaConf.create(cfg.dump())
if not torch.cuda.is_available():
cfg.MODEL.DEVICE = "cpu"
# test that a model can be built with omegaconf config as well
build_model(cfg)
def test_setup_config(self):
with tempfile.TemporaryDirectory(prefix="detectron2_test") as d:
cfg = get_cfg()
cfg.OUTPUT_DIR = os.path.join(d, "yacs")
default_setup(cfg, {})
cfg = model_zoo.get_config(
"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.py")
cfg.train.output_dir = os.path.join(d, "omegaconf")
default_setup(cfg, {})
def test_fpn_scriptability(self):
cfg = model_zoo.get_config(
"Misc/scratch_mask_rcnn_R_50_FPN_3x_gn.yaml")
bb = build_resnet_fpn_backbone(cfg, ShapeSpec(channels=3))
bb_s = torch.jit.script(bb)
inp = torch.rand(2, 3, 128, 128)
out1 = bb(inp)["p5"]
out2 = bb_s(inp)["p5"]
self.assertTrue(torch.allclose(out1, out2))
def get_model_no_weights(config_path):
"""
Like model_zoo.get, but do not load any weights (even pretrained)
"""
cfg = model_zoo.get_config(config_path)
if isinstance(cfg, CfgNode):
if not torch.cuda.is_available():
cfg.MODEL.DEVICE = "cpu"
return build_model(cfg)
else:
return instantiate(cfg.model)
def test_PointRend_mask_head_tracing(self):
cfg = model_zoo.get_config(
"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml")
point_rend.add_pointrend_config(cfg)
cfg.MODEL.ROI_HEADS.IN_FEATURES = ["p2", "p3"]
cfg.MODEL.ROI_MASK_HEAD.NAME = "PointRendMaskHead"
cfg.MODEL.ROI_MASK_HEAD.POOLER_TYPE = ""
cfg.MODEL.ROI_MASK_HEAD.POINT_HEAD_ON = True
chan = 256
head = point_rend.PointRendMaskHead(
cfg,
{
"p2": ShapeSpec(channels=chan, stride=4),
"p3": ShapeSpec(channels=chan, stride=8),
},
)
def gen_inputs(h, w, N):
p2 = torch.rand(1, chan, h, w)
p3 = torch.rand(1, chan, h // 2, w // 2)
boxes = random_boxes(N, max_coord=h)
return p2, p3, boxes
class Wrap(nn.ModuleDict):
def forward(self, p2, p3, boxes):
features = {
"p2": p2,
"p3": p3,
}
inst = Instances((p2.shape[2] * 4, p2.shape[3] * 4))
inst.pred_boxes = Boxes(boxes)
inst.pred_classes = torch.zeros(inst.__len__(),
dtype=torch.long)
out = self.head(features, [inst])[0]
return out.pred_masks
model = Wrap({"head": head})
model.eval()
with torch.no_grad(), patch_builtin_len():
traced = torch.jit.trace(model, gen_inputs(302, 208, 20))
inputs = gen_inputs(100, 120, 30)
out_eager = model(*inputs)
out_trace = traced(*inputs)
self.assertTrue(torch.allclose(out_eager, out_trace))
def init_context(context):
context.logger.info("Init context... 0%")
cfg = get_config('COCO-Detection/retinanet_R_101_FPN_3x.yaml')
if torch.cuda.is_available():
CONFIG_OPTS.extend(['MODEL.DEVICE', 'cuda'])
else:
CONFIG_OPTS.extend(['MODEL.DEVICE', 'cpu'])
cfg.merge_from_list(CONFIG_OPTS)
cfg.MODEL.RETINANET.SCORE_THRESH_TEST = CONFIDENCE_THRESHOLD
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = CONFIDENCE_THRESHOLD
cfg.MODEL.PANOPTIC_FPN.COMBINE.INSTANCES_CONFIDENCE_THRESH = CONFIDENCE_THRESHOLD
cfg.freeze()
predictor = DefaultPredictor(cfg)
context.user_data.model_handler = predictor
context.logger.info("Init context...100%")
def _test_model(self, config_path, device="cpu"):
# requires extra dependencies
from detectron2.export import Caffe2Model, add_export_config, Caffe2Tracer
cfg = model_zoo.get_config(config_path)
add_export_config(cfg)
cfg.MODEL.DEVICE = device
model = model_zoo.get(config_path, trained=True, device=device)
inputs = [{"image": get_sample_coco_image()}]
c2_model = Caffe2Tracer(cfg, model,
copy.deepcopy(inputs)).export_caffe2()
with tempfile.TemporaryDirectory(prefix="detectron2_unittest") as d:
c2_model.save_protobuf(d)
c2_model.save_graph(os.path.join(d, "test.svg"),
inputs=copy.deepcopy(inputs))
c2_model = Caffe2Model.load_protobuf(d)
c2_model(inputs)[0]["instances"]
def _test_model(self, config_path, device="cpu"):
cfg = model_zoo.get_config(config_path)
cfg.MODEL.DEVICE = device
model = model_zoo.get(config_path, trained=True, device=device)
inputs = [{"image": get_sample_coco_image()}]
tracer = Caffe2Tracer(cfg, model, copy.deepcopy(inputs))
with tempfile.TemporaryDirectory(prefix="detectron2_unittest") as d:
if not os.environ.get("CI"):
# This requires onnx, which is not yet available on public CI
c2_model = tracer.export_caffe2()
c2_model.save_protobuf(d)
c2_model.save_graph(os.path.join(d, "test.svg"),
inputs=copy.deepcopy(inputs))
c2_model = Caffe2Model.load_protobuf(d)
c2_model(inputs)[0]["instances"]
ts_model = tracer.export_torchscript()
ts_model.save(os.path.join(d, "model.ts"))
from detectron2.model_zoo import get_config
model = get_config("common/models/retinanet.py").model
model.backbone.bottom_up.freeze_at = 2
model.head.norm = "SyncBN"
dataloader = get_config("common/data/coco.py").dataloader
lr_multiplier = get_config("common/coco_schedule.py").lr_multiplier_3x
optimizer = get_config("common/optim.py").SGD
train = get_config("common/train.py").train
optimizer.lr = 0.01
train.init_checkpoint = "detectron2://ImageNetPretrained/MSRA/R-50.pkl"
train.max_iter = 270000 # 3x for batchsize = 16
std=(0.229, 0.224, 0.225)),
]),
),
batch_size=256 // 8,
num_workers=4,
training=False,
)
dataloader.evaluator = L(ClassificationAcc)()
model = L(ClassificationNet)(model=(
ResNet)(block=Bottleneck, layers=[3, 4, 6, 3], zero_init_residual=True))
optimizer = L(torch.optim.SGD)(
params=L(get_default_optimizer_params)(),
lr=0.1,
momentum=0.9,
weight_decay=1e-4,
)
lr_multiplier = L(WarmupParamScheduler)(
scheduler=L(MultiStepParamScheduler)(values=[1.0, 0.1, 0.01, 0.001],
milestones=[30, 60, 90, 100]),
warmup_length=1 / 100,
warmup_factor=0.1,
)
train = get_config("common/train.py").train
train.init_checkpoint = None
train.max_iter = 100 * 1281167 // 256
parser.add_argument("--test_annotations", type=str, default='/home/dimitar/test_annotations.json', help="Path to COCO-style annotations file for model evaluation.")
parser.add_argument("--test_imagedir", type=str, default='/home/mengmi/Projects/Proj_context2/Matlab/Stimulus/keyframe_expH', help="Path to images folder w.r.t. which filenames are specified in the annotations for model evaluation.")
parser.add_argument("--test_frequency", type=int, default=50000, help="Evaluate model on test data every __ iterations.")
parser.add_argument("--iters", type=int, default=500000, help="Number of iterations to train.")
parser.add_argument("--save_frequency", type=int, default=50000, help="Save model checkpoint every __ iterations.")
args = parser.parse_args()
# Register datasets
register_coco_instances("train", {}, args.annotations, args.imagedir)
register_coco_instances("test", {}, args.test_annotations, args.test_imagedir)
# Load and configure model
cfg = model_zoo.get_config("COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml", trained=True)
cfg.DATASETS.TRAIN = ("train",)
cfg.DATASETS.TEST = ("test",)
cfg.OUTPUT_DIR = args.outdir
cfg.DATALOADER.NUM_WORKERS = 4
cfg.SOLVER.IMS_PER_BATCH = 2
cfg.SOLVER.BASE_LR = 0.00025
cfg.SOLVER.CHECKPOINT_PERIOD = args.save_frequency
cfg.SOLVER.MAX_ITER = args.iters # Note that when traininig is resumed the iteration count will resume as well, so increase the number of iterations to train further.
cfg.TEST.EVAL_PERIOD = args.test_frequency
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 55
# save config
