treecounter-ML VS multi-object-tracker

You can take a look at an example implementation of the code here. I was wondering how I can extract the feature maps from a particular layer using the dnn module. I thought of using the net.forward() to do it, but I am not sure if this is the correct way, and more importantly, I do not want to run two forward passes on the same image(I do not think it will be an elegant implementation).

Overlapping Bounding Boxes: As you can see from the image, I have quite a lot of overlapping bounding boxes around each detected object. I did some reading online, and found that I needed to use Non Maxima Suppression to filter the detections. I did that, and adjusted the NMS_THRESH parameter(The current results are at confidence threshold of 95% and NMS threshold of 100%). I am perplexed by this, and what I would like to have is some sort of a filter, where say if I already have a box at (x, y, w, h), then there should be like a barrier around the surrounding n pixels, where even if a box is detected, it will not be placed. If you want to take a look at my code, you can find the repository here and the particular file here.

I am not being able to work the Sort tracker into my current application code. It would be great if any of you could help me out here. The program crashes when I use the tracker.update() method.

You may find this useful: https://adipandas.github.io/multi-object-tracker/

As you can see, in Frame 43, the tracker assigns the ID 10 to a metal post, but a few frames later when the tracker tracks the same metal post, it provides an ID of 11. This shows that the tracker is assigning new IDs to the same detected object. I cannot seem to find out why this happens, and how to fix it. I am using the motrackers module from this repo. I have not made any changes as such in the mot_yolov3.py file, I have just added the line to print the frame counter.

I have a simple python project which has gone through at least 5 iterations since I began working on it. Please see this link: adipandas/multi-object-tracker.