zongdai/AutoShape

no visualization effect but only original image

Light-- opened this issue · 1 comments

after run:

python src/faster.py --demo /home/kitti/testing/image_2 --calib_dir /home/kitti/testing/calib --load_model trainval_model_batch16_gpu2_200epoch_pc48_adaptivekp_rightaug.pth --gpus 0 --arch dla_34 --num_joints 57 --sample_pc 48 --vis

i only got empty white bev and original image

image

not as expected

Hi, I also tried to visualize it by myself and find the way like this:
You can add this line :

debugger.save_detection_img(bbox, self.image_path, self.opt, calib, img_id='car_pose')

in the function show_results(self, debugger, image, results, calib) of pytorch/src/lib/detectors/auto_shape.py file.
Like this:
image