Evaluate results on YoloFastest V2 (.pth model)

HaiderAbasi / OpenCV-Raspberry-Pi-4-Projects-on-the-Edge

Harnessing the power of Raspberry Pi 4 to build cutting-edge computer vision solutions. Whether you're interested in object detection, image classification, or real-time video analysis, this project will give you the tools you need to get started.

MIT License

0 stars 0 forks source link

Evaluate results on YoloFastest V2 (.pth model) #3

Closed HaiderAbasi closed 1 year ago

HaiderAbasi commented 1 year ago

Testing on Frames detected only cars.

Detected objects are >90% correct (low FP) :)
Missing alot of other vehicles > (high FN) :(
Smaller objects are harder to detect.

Further testing on videos is required to confirm.

Inference speed on CPU.
Accuracy on multiple vehicles in a fast-changing environment.

HaiderAbasi commented 1 year ago

Testing on YoloFastest-V2 (,pth model)

Testing on Frames > done <
Testing on test-videos. (Awaiting results)

HaiderAbasi commented 1 year ago

Testing on YoloFastest-V2 (,pth model)

Testing on Frames > done <
Testing on test videos. > done <

Results of the test videos

Varying object classifications - Might need to restrict our goals 2, Multiple Overlapping predictions - IOU adjustment needed
Ghost detections - restrict to focus on road only