search

Pandas-Team / Autonomous-Vehicle-Environment-Perception

An Intelligent Modular Real-Time Vision-Based System for Environment Perception (NeurIPS 2022 Workshop)
MIT License
67 stars 23 forks source link

Green area. #1

Closed MosheMikhael closed 3 years ago

MosheMikhael commented 3 years ago

Good day! Beautiful project, I'm trying to run in сolab on my video and get unsatisfactory results, I guess because of the low quality of the picture + bad road markings, isn't it? What does the green area on the picture mean? image

miladsoltany commented 3 years ago

Dear Moshe, Thank you for your interest in our project. The green area indicates crosswalks on the street. We used classic computer vision algorithms used for this particular task, and the thresholds are hand-chosen. Therefore, it's not robust against low-quality images. We are, however, trying to fix that in near future.

MosheMikhael commented 3 years ago

One little thing, notice that in PINet.py, in the description of the LaneDetection() class, in the Testing(self, frame, mask) function there is an unpleasant line: ti[0] = cv2.resize(ti[0], (1280,720)). If we run on a video with a different size, there will be an AssertationError somewhere in the mask overlay process.

Good luck!

miladsoltany commented 3 years ago

@MosheMikhael Thanks for noticing the issue and telling us about it, we will soon work on fixing the issues you mentioned.

miladsoltany commented 3 years ago

Dear @MosheMikhael, I added a quick fix to the assertion error problem you mentioned. Thank you for your feedback.