def mots_challenge_train():
return get_kiti_mots_dicts(ims_path,
annots_path,
is_train=True,
train_percentage=1.,
image_extension='jpg')
MetadataCatalog.get(cfg.DATASETS.TRAIN[0]).set(
crop=opts.crop, hflip=opts.hflip, change_contrast=opts.contrast)
cfg.OUTPUT_DIR = output_dir
os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)
trainer = TrainerDA(cfg)
val_loss = ValidationLoss(cfg)
trainer.register_hooks([val_loss])
trainer._hooks = trainer._hooks[:-2] + trainer._hooks[-2:][::-1]
trainer.resume_or_load(resume=True)
trainer.train()
if not opts.train_only:
evaluator = COCOEvaluator("kitti_mots_test",
cfg,
False,
output_dir=output_dir)
trainer.test(cfg, trainer.model, evaluators=[evaluator])
plot_losses(cfg)
predictor = DefaultPredictor(cfg)
predictor.model.load_state_dict(trainer.model.state_dict())
dataset_dicts = get_kiti_mots_dicts(
"../datasets/KITTI-MOTS/training/image_02",
"../datasets/KITTI-MOTS/instances_txt",
is_train=False,
image_extension='png')
show_results(cfg, dataset_dicts, predictor, samples=10)
def kitti_mots_test():
return get_kiti_mots_dicts(ims_path_kitti,
annots_path_kitti,
is_train=False,
train_percentage=train_percent_kitti_mots,
image_extension='png')