Training coco part code is missing.

[TyroneLi]

In folder step, train_cam_coco.py is missing. Can you upload it? Thans

[TyroneLi]

By the way, your code for refining the initial cam see using CRF is not yet released.

[TyroneLi]

Hello~,anybody~

[jbeomlee93]

Sorry for the late reply.

1. I have updated "step/train_cam_coco.py", but it has not been newly tested.

You can run RIB process with pre-trained classifier provided by https://drive.google.com/file/d/1SDKjPzzuXR4PX_H1l3Kw5NY8d20FmG4X/view.

2. You can refine the initial cam using CRF similar to below.

RIB/step/cam_to_ir_label.py

Lines 21 to 35 in 9c43b3b

	cam_dict = np.load(os.path.join(args.cam_out_dir, img_name + '.npy'), allow_pickle=True).item()
	cams = cam_dict['high_res']
	keys = np.pad(cam_dict['keys'] + 1, (1, 0), mode='constant')
	# 1. find confident fg & bg
	fg_conf_cam = np.pad(cams, ((1, 0), (0, 0), (0, 0)), mode='constant', constant_values=args.conf_fg_thres)
	fg_conf_cam = np.argmax(fg_conf_cam, axis=0)
	pred = imutils.crf_inference_label(img, fg_conf_cam, n_labels=keys.shape[0])
	fg_conf = keys[pred]
	bg_conf_cam = np.pad(cams, ((1, 0), (0, 0), (0, 0)), mode='constant', constant_values=args.conf_bg_thres)
	bg_conf_cam = np.argmax(bg_conf_cam, axis=0)
	pred = imutils.crf_inference_label(img, bg_conf_cam, n_labels=keys.shape[0])
	bg_conf = keys[pred]