Lightweight Python library for adding real-time object tracking to any detector.

Hi,

I'm trying to implement a tracker which uses embedding. I'm storing each embedding in the detection.data, however I also need a place where to store the TrackedObject embedding, in order to compute a suitable distance. Is there something similar to detection.data for TrackedObject?

Hello ,

I am trying to implement norfair along with yolov4 with my own dataset. I want to know how well norfair is tracking objects in my dataset ? can you please help me in evaluating norfair ?

Hello, thanks for sharing this awesome tracker. Im trying to integrate it to darknet, but its not behaving as expected.

heres what i do:

detections = darknet.detect_image(network, class_names, darknet_image, thresh=0.2)
        detections2 = [
            Detection(get_centroid(detection[2], width, height), data=detection[2])
            for detection in detections
        ]
tracked_objects = tracker.update(detections=detections2)
norfair.draw_tracked_objects(frame_resized, tracked_objects)

the tracker is somewhat populated if i do a print(tracker.tracker_objects) but it doesnt look right and the draw_tracked_objects is dead. since darknet outputs x and y by default ive tried a variety of variations with converting to bbox etc. nothing has worked and im hoping for a little push :)

is the data parameter necessary in my case?

edit i forgot to output darknets centroid as numpy array, its working now :) for those who maybe end up in the same issue, change out get_centroid with something like this (just clean up my messy demodef):

def get_detlist(detections):
    arr = np.empty((0, 2), int)
    arr = np.append(arr, np.array([[detections[0],detections[1]]]), axis=0)
    return arr
    
Detection(get_detlist(detection[2]), data=detection[2])

Demos:

• [x] 3d_track
  • already uses videos
• [x] alphapose
• [x] camera_motion
• [x] detectron2
• [x] keypoints_bounding_boxes
• [x] mmdetection
• [x] motmetrics4norfair
• [x] openpose
  • skipped 1/2 frames instead od 4/5 because it didn't work well
• [x] profiling
• [x] reid
  • had to change the table for a single video because tables don't work with mp4
• [x] yolov4
• [x] yolov5
• [x] yolov7
  • updated the demo to use bboxes by default

closes #169 and #167

Hi,

I am trying to track pose estimates using the "Tracking pedestrians with AlphaPose" demo as a reference. However I am using Nvidia trt-pose (https://github.com/NVIDIA-AI-IOT/trt_pose) instead of alpha pose as given in the demo.

The pose estimation alone runs well at around 25fps (having about 50% CPU usage), however when I include the pose tracking, my fps drops to about 10-12 fps and its definitely a CPU bottleneck as my CPU usage is around 98% when running tracking. I would like to know if this is considered "normal" with the pose estimation tracking or I am doing something wrong in my end.

PC specs GTX 1060 6GB intel i7 8500 H 6GB ram

Thanks for the great work.

Hi & Thanks for your wonderful library.

I run your tracker with "reid" feature enabled and it seems like there is a small bug in calculating "reid_hit_counter". (Although it seems too obvious so probably I am missing a point)

As the document stated :

Each tracked object keeps an internal ReID hit counter which tracks how often it's getting recognized by another tracker, each time it gets a match this counter goes up, and each time it doesn't it goes down. If it goes below 0 the object gets destroyed. If used, this argument (reid_hit_counter_max) defines how long an object can live without getting matched to any detections, before it is destroyed.

But there is no code for increasing the value of 'reid_hit_counter'. The only part of the code which I found is in tracker_step :

    def tracker_step(self):
        self.hit_counter -= 1
        if self.reid_hit_counter is None:
            if self.hit_counter <= 0:
                self.reid_hit_counter = self.reid_hit_counter_max
        else:
            self.reid_hit_counter -= 1
        self.point_hit_counter -= 1
        self.age += 1
        # Advances the tracker's state
        self.filter.predict()

So even if the object is "alive", reid_hit_counter keeps decreasing and as soon as it gets to zero, according to this section of the code :

if self.reid_hit_counter_max is None:
            self.tracked_objects = [
                o for o in self.tracked_objects if o.hit_counter_is_positive
            ]
            alive_objects = self.tracked_objects
        else:
            tracked_objects = []
            for o in self.tracked_objects:
                if o.reid_hit_counter_is_positive:
                    tracked_objects.append(o)
                    if o.hit_counter_is_positive:
                        alive_objects.append(o)
                    else:
                        dead_objects.append(o)

that object gets suddenly destroyed. It happens mostly to a long living objects in the scene which reid_hit_counter has time to get zero. Looks like the "alive" objects should not get their reid_hit_counter decreased.

Thanks in advance

I'm using Norfair and another library. However, I'm getting a version conflict for rich.

Is it possible to change the dependency of norfair to be able to use higher version of rich? I'm using poetry to install norfair and I'm getting a "SolveProblemError Because norfair (0.4.0) depends on rich (>=9.10.0,<10.0.0)". I need to use at least version 11.2.0.

I can force install higher version of rich using pip but I think it would be cleaner if the dependencies are clear. Norfair seems to work fine even with higher version of rich.,

@gerhc @draix @dekked hi thanks for sharing this source code when i try to install pip install norfair[video] # Adds several video helper features running on OpenCV pip install norfair[metrics] # Supports running MOT metrics evaluation pip install norfair[metrics,video] # Everything included I i get the following error "ImportError: cannot import name 'metrics'"

Please let me knw what is the issue

Thanks for this great library :+1:

I'm wondering why is here the prediction score of the detection is exported as -1? https://github.com/tryolabs/norfair/blob/4d59a4d5dd0cf4738c70de6e93782c93bce10d41/norfair/metrics.py#L93

Are there any more examples of distance functions? I'm doing tracking with bounding boxes, and a feature extractor on the box. Is there an example that uses this type of setup for norfair?

Hi, I have a question related to issue # 19: I'm working through the detectron2 demo (detectron2_cars.py) to understand the code and wanted to change the detected class from cars to person. I found that changing row 32 to "if c == 0" does this. How do I export/print a list of the detected classes?

Also, I would like to export to a csv file the centroid/center coordinates for each tracked object for each frame (similar to print_objects_as_table, but ideally with the data from each frame in one row). Any suggestions?

im using Yolov5 to detect objects on my screen, iam using mss to take screenshots and feed it to yolov5, how can i implent norfair with a setup like this?

Describe the situation you are working on I would like to thank you for this amazing git ..I am using yolov5 model detection with the norfair tracker..

Describe what is it that you need help with In the part where we want to draw the bounding boxes ..I initially used the function draw_tracked_boxes but got the message that this function deprecated..now to use the draw_box function I am not sure how input should be given should I pass the detections of yolov5 or should I pass tracked_objects Additional context

I saw the part where an argument drawables is used it is said that this should be union of detections and tracked_objects but the detection object is this <tracker.Detection object at 0x7f7f7580c2e8> and tracked_object is Object_7(age: 7, hit_counter: 8, last_distance: 0.62, init_id: 1)

Not sure what can be done..

Until now, the default value for past_detections_length was 4. This meant that even if you don't use the past detections in your workflow, Norfair is still storing them, making unnecessary operations and potentially using a great amount of memory for storing embeddings or image crops.

We now set the default value of past_detections_length to 0. If the user wants to use the TrackedObject's past detections when matching, it must be specified when initiating the Tracker instance.