maltreba/norfair

Norfair is a customizable lightweight Python library for real-time 2D object tracking.

Using Norfair, you can add tracking capabilities to any detector with just a few lines of code.

Features

Any detector expressing its detections as a series of (x, y) coordinates can be used with Norfair. This includes detectors performing object detection, pose estimation, and interest point detection.
The function used to calculate the distance between tracked objects and detections is defined by the user, making the tracker extremely customizable. This function can make use of any extra information, such as appearance embeddings, which can heavily improve tracking performance.
Modular. It can easily be inserted into complex video processing pipelines to add tracking to existing projects. At the same time it is possible to build a video inference loop from scratch using just Norfair and a detector.
Fast. The only thing bounding inference speed will be the detection network feeding detections to Norfair.

Norfair is built, used and maintained by Tryolabs.

Installation

Norfair currently supports Python 3.6+.

For the minimal version, install as:

pip install norfair

To make Norfair install the dependencies to support more features, install as:

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

If the needed dependencies are already present in the system, installing the minimal version of Norfair is enough for enabling the extra features. This is particuarly useful for embedded devices, where installing compiled dependencies can be difficult, but they can sometimes come preinstalled with the system.

How it works

Norfair works by estimating the future position of each point based on its past positions. It then tries to match these estimated positions with newly detected points provided by the detector. For this matching to occur, Norfair can rely on any distance function specified by the user of the library. Therefore, each object tracker can be made as simple or as complex as needed.

The following is an example of a particularly simple distance function calculating the Euclidean distance between tracked objects and detections. This is possibly the simplest distance function you could use in Norfair, as it uses just one single point per detection/object.

 def euclidean_distance(detection, tracked_object):
     return np.linalg.norm(detection.points - tracked_object.estimate)

As an example we use Detectron2 to get the single point detections to use with this distance function. We just use the centroids of the bounding boxes it produces around cars as our detections, and get the following results.

On the left you can see the points we get from Detectron2, and on the right how Norfair tracks them assigning a unique identifier through time. Even a straightforward distance function like this one can work when the tracking needed is simple.

Norfair also provides several useful tools for creating a video inference loop. Here is what the full code for creating the previous example looks like, including the code needed to set up Detectron2:

import cv2
import numpy as np
from detectron2.config import get_cfg
from detectron2.engine import DefaultPredictor

from norfair import Detection, Tracker, Video, draw_tracked_objects

# Set up Detectron2 object detector
cfg = get_cfg()
cfg.merge_from_file("demos/faster_rcnn_R_50_FPN_3x.yaml")
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5
cfg.MODEL.WEIGHTS = "detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl"
detector = DefaultPredictor(cfg)

# Norfair
video = Video(input_path="video.mp4")
tracker = Tracker(distance_function=euclidean_distance, distance_threshold=20)

for frame in video:
    detections = detector(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
    detections = [Detection(p) for p in detections['instances'].pred_boxes.get_centers().cpu().numpy()]
    tracked_objects = tracker.update(detections=detections)
    draw_tracked_objects(frame, tracked_objects)
    video.write(frame)

The video and drawing tools use OpenCV frames, so they are compatible with most Python video code available online. The point tracking is based on SORT generalized to detections consisting of a dynamically changing amount of points per detection.

Motivation

Trying out the latest state of the art detectors normally requires running repositories which weren't intended to be easy to use. These tend to be repositories associated with a research paper describing a novel new way of doing detection, and they are therefore intended to be run as a one-off evaluation script to get some result metric to publish on a particular research paper. This explains why they tend to not be easy to run as inference scripts, or why extracting the core model to use in another standalone script isn't always trivial.

Norfair was born out of the need to quickly add a simple layer of tracking over a wide range of newly released SOTA detectors. It was designed to seamlessly be plugged into a complex, highly coupled code base, with minimum effort. Norfair provides a series of modular but compatible tools, which you can pick and chose to use in your project.

Examples

We provide several examples of how Norfair can be used to add tracking capabilities to several different detectors.

Simple tracking of cars using Detectron2.
Faster tracking of cars/pedestrians and other 78 classes using YOLOv5. Try it on any YouTube video on this Google Colab notebook.
Faster tracking of cars using YOLOv4.
Inserting Norfair into an existing project: Simple tracking of pedestrians using AlphaPose.
Speed up pose estimation by extrapolating detections using OpenPose.
Tracking objects from different classes using keypoints and bounding boxes using YOLOv5 and OpenPose.
Filter and distance function profiling using TRT pose estimator.

Comparison to other trackers

Norfair's contribution to Python's object tracker library repertoire is its ability to work with any object detector by being able to work with a variable number of points per detection, and the ability for the user to heavily customize the tracker by creating their own distance function.

If you are looking for a tracker, here are some other projects worth noting:

OpenCV includes several tracking solutions like KCF Tracker and MedianFlow Tracker which are run by making the user select a part of the frame to track, and then letting the tracker follow that area. They tend not to be run on top of a detector and are not very robust.
dlib includes a correlation single object tracker. You have to create your own multiple object tracker on top of it yourself if you want to track multiple objects with it.
AlphaPose just released a new version of their human pose tracker. This tracker is tightly integrated into their code base, and to the task of tracking human poses.
SORT and Deep SORT are similar to this repo in that they use Kalman filters (and a deep embedding for Deep SORT), but they are hardcoded to a fixed distance function and to tracking boxes. Norfair also adds some filtering when matching tracked objects with detections, and changes the Hungarian Algorithm for its own distance minimizer. Both these repos are also released under the GPL license, which might be an issue for some individuals or companies because the source code of derivative works needs to be published.

Benchmarks

MOT17 results obtained using motmetrics4norfair demo script. Hyperparameters were tuned for reaching a high MOTA on this dataset. A more balanced set of hyperparameters, like the default ones used in the other demos, is recommended for production.

	Rcll	Prcn	GT MT PT ML	FP FN IDs FM	MOTA MOTP
MOT17-13-DPM	18.0%	83.5%	110 5 28 77	416 9543 120 125	13.4% 26.8%
MOT17-04-FRCNN	56.3%	93.2%	83 18 43 22	1962 20778 90 104	52.0% 10.7%
MOT17-11-FRCNN	61.5%	93.1%	75 19 33 23	431 3631 64 61	56.3% 10.1%
MOT17-04-SDP	77.6%	97.4%	83 48 26 9	1001 10672 225 254	75.0% 13.2%
MOT17-13-SDP	57.0%	82.1%	110 45 28 37	1444 5008 229 209	42.6% 20.1%
MOT17-05-DPM	38.0%	82.2%	133 10 58 65	570 4291 96 96	28.3% 24.2%
MOT17-09-DPM	59.9%	75.4%	26 4 18 4	1042 2137 119 113	38.1% 26.2%
MOT17-10-DPM	37.3%	84.6%	57 6 20 31	871 8051 127 154	29.5% 24.8%
MOT17-02-SDP	51.0%	76.1%	62 11 39 12	2979 9103 268 284	33.5% 18.2%
MOT17-11-DPM	54.2%	84.5%	75 12 24 39	935 4321 88 64	43.4% 21.7%
MOT17-09-FRCNN	58.6%	98.5%	26 7 17 2	49 2207 40 39	56.9% 9.5%
MOT17-11-SDP	75.8%	91.1%	75 34 30 11	697 2285 103 101	67.3% 14.0%
MOT17-02-FRCNN	36.6%	79.7%	62 7 26 29	1736 11783 119 131	26.6% 13.4%
MOT17-05-FRCNN	54.7%	89.2%	133 24 68 41	457 3136 95 96	46.7% 18.1%
MOT17-04-DPM	42.5%	83.6%	83 7 44 32	3965 27336 401 420	33.3% 21.0%
MOT17-10-SDP	74.2%	88.1%	57 30 24 3	1292 3316 308 289	61.7% 19.8%
MOT17-10-FRCNN	61.0%	75.8%	57 16 35 6	2504 5013 319 313	39.0% 17.3%
MOT17-09-SDP	67.6%	94.6%	26 12 14 0	204 1726 52 55	62.8% 13.0%
MOT17-02-DPM	20.2%	81.6%	62 5 14 43	843 14834 111 112	15.0% 24.6%
MOT17-13-FRCNN	58.8%	73.6%	110 29 57 24	2455 4802 371 366	34.5% 18.5%
MOT17-05-SDP	66.7%	87.6%	133 32 81 20	653 2301 134 140	55.4% 16.5%
OVERALL	53.6%	87.2%	1638 381 727 530	26506 156274 3479 3526	44.7% 16.3%

Commercial support

Tryolabs can provide commercial support, implement new features in Norfair or build video analytics tools for solving your challenging problems. Norfair powers several video analytics applications, such as the face mask detection tool.

If you are interested, please contact us.

Citing Norfair

For citations in academic publications, please export your desired citation format (BibTeX or other) from Zenodo.

maltreba / norfair