Could not achieve same results in demo

[tszhang97]

I use the code of demo.py and test the video in the demo. The results below look wired. I also test other videos and a common problem is that the results have lots of jitters.
https://github.com/thohemp/6DRepNet/assets/39046939/740edd75-7565-4bfa-a82b-0303b7ee9bde

[tszhang97]

video results:
https://github.com/thohemp/6DRepNet/assets/39046939/8f45a2ed-7e46-4a20-bbde-446bea43a3aa

[thohemp]

Did you use my demo video with box visualization as input to process it again?

[tszhang97]

@thohemp Do you mean this part？ I've already used the code below:
detector = RetinaFace(gpu_id=gpu)
with torch.no_grad():
for frame in img_list:
faces = detector(frame)
for box, landmarks, score in faces:
# Print the location of each face in this image
if score < .95:
continue
x_min = int(box[0])
y_min = int(box[1])
x_max = int(box[2])
y_max = int(box[3])

            bbox_width = abs(x_max - x_min)
            bbox_height = abs(y_max - y_min)

            x_min = max(0, x_min - int(0.2 * bbox_height))
            y_min = max(0, y_min - int(0.2 * bbox_width))
            x_max = x_max + int(0.2 * bbox_height)
            y_max = y_max + int(0.2 * bbox_width)
            img = frame[y_min:y_max, x_min:x_max]