Test on my own image

Open TienDungHUST opened this issue 3 years ago • 1 comments

I have run demo-evalScaleNet-COCOScale-RELEASE.ipynb successfully. While the person's height estimation on COCO make sense, I realize the height estimation of car on COCO val-set is not good. Therefore, I want to inference on my own image which contain the car/truck with only known 2D bbox. Since the COCO's preprocess too complex, how can I replace the image input?

Jul 21 '22 02:07 TienDungHUST

I have run demo-evalScaleNet-COCOScale-RELEASE.ipynb successfully. While the person's height estimation on COCO make sense, I realize the height estimation of car on COCO val-set is not good. Therefore, I want to inference on my own image which contain the car/truck with only known 2D bbox. Since the COCO's preprocess too complex, how can I replace the image input?

May I ask how did you successfully run the demo-evalScaleNet-COCOScale-RELEASE.ipynb? I can't find the module named 'debugging' in the source code in the File "ScaleNet-master\RELEASE_ScaleNet_minimal\dataset_cvpr.py", line 18, in from debugging import getHorizonLine

Apr 18 '24 07:04 Aserck-ck