|
1 | | -# IOU-Tracker |
| 1 | +## IOU Tracker |
| 2 | +Python implementation of the IOU Tracker described in the AVSS 2017 paper |
| 3 | +[High-Speed Tracking-by-Detection Without Using Image Information](http://elvera.nue.tu-berlin.de/files/1517Bochinski2017.pdf). |
2 | 4 |
|
3 | | -coming soon... |
| 5 | +This project is released under the MIT License (details in LICENSE file). |
| 6 | +If you think our work is useful in your research, please consider citing: |
| 7 | + |
| 8 | +``` |
| 9 | +@INPROCEEDINGS{1517Bochinski2017, |
| 10 | + AUTHOR = {Erik Bochinski and Volker Eiselein and Thomas Sikora}, |
| 11 | + TITLE = {High-Speed Tracking-by-Detection Without Using Image Information}, |
| 12 | + BOOKTITLE = {International Workshop on Traffic and Street Surveillance for Safety and Security at IEEE AVSS 2017}, |
| 13 | + YEAR = {2017}, |
| 14 | + MONTH = aug, |
| 15 | + ADDRESS = {Lecce, Italy}, |
| 16 | + URL = {http://elvera.nue.tu-berlin.de/files/1517Bochinski2017.pdf}, |
| 17 | + } |
| 18 | +``` |
| 19 | + |
| 20 | +## Demo |
| 21 | +Several demo scripts are included to reproduce the reported results on the UA-DETRAC |
| 22 | +and the MOT benchmark. |
| 23 | + |
| 24 | +Basic demo script: |
| 25 | +``` |
| 26 | +$ ./demo.py -h |
| 27 | +usage: demo.py [-h] -d DETECTION_PATH -o OUTPUT_PATH [-sl SIGMA_L] |
| 28 | + [-sh SIGMA_H] [-si SIGMA_IOU] [-tm T_MIN] |
| 29 | +
|
| 30 | +IOU Tracker demo script |
| 31 | +
|
| 32 | +optional arguments: |
| 33 | + -h, --help show this help message and exit |
| 34 | + -d DETECTION_PATH, --detection_path DETECTION_PATH |
| 35 | + full path to CSV file containing the detections |
| 36 | + -o OUTPUT_PATH, --output_path OUTPUT_PATH |
| 37 | + output path to store the tracking results (MOT |
| 38 | + challenge devkit compatible format) |
| 39 | + -sl SIGMA_L, --sigma_l SIGMA_L |
| 40 | + low detection threshold |
| 41 | + -sh SIGMA_H, --sigma_h SIGMA_H |
| 42 | + high detection threshold |
| 43 | + -si SIGMA_IOU, --sigma_iou SIGMA_IOU |
| 44 | + intersection-over-union threshold |
| 45 | + -tm T_MIN, --t_min T_MIN |
| 46 | + minimum track length |
| 47 | +``` |
| 48 | + |
| 49 | +Example: |
| 50 | +``` |
| 51 | +./demo.py -d ../mot17/train/MOT17-04-SDP/det/det.txt -o res/iou-tracker/MOT17-04-SDP.txt |
| 52 | +``` |
| 53 | + |
| 54 | +### DETRAC |
| 55 | +To reproduce the reported results, download and extract the [DETRAC-toolkit](http://detrac-db.rit.albany.edu/download) |
| 56 | +and the detections you want to evaluate. Download links for the EB detections are provided below. |
| 57 | +Clone this repository into "DETRAC-MOT-toolkit/trackers/". |
| 58 | +Follow the instructions to configure the toolkit for tracking evaluation and set the tracker name in "DETRAC_experiment.m": |
| 59 | + |
| 60 | +``` |
| 61 | +tracker.trackerName = 'iou-tracker'; |
| 62 | +``` |
| 63 | + |
| 64 | +and run the script. |
| 65 | + |
| 66 | +Note that you still need a working python environment. |
| 67 | +You should obtain something like the following results for the 'DETRAC-Train' set: |
| 68 | + |
| 69 | +##### DETRAC-Train Results |
| 70 | +| Detector | Rcll | Prcn | FAR | MT | PT | ML | FP | FN | IDs | FM | MOTA | MOTP | MOTAL | |
| 71 | +| -------- | ---- | ---- | --- | ----- | --- | ------ | ------- | ------- | ----- | ----- |----- |------|-------| |
| 72 | +| EB |22.12 |31.53 |0.26 |17.65 |13.22|18.41 |14796.52 |171806.84|2311.25|2445.89|19.41 |28.89 |19.78 | |
| 73 | +| R-CNN |27.86 |52.90 |0.11 |19.53 |17.03|18.56 |9047.95 |157521.18|4842.18|4969.57|25.46 |44.39 |26.29 | |
| 74 | + |
| 75 | +##### DETRAC-Test (Overall) Results |
| 76 | +| Detector | Rcll | Prcn | FAR | MT | PT | ML | FP | FN | IDs | FM | MOTA | MOTP | MOTAL | |
| 77 | +| -------- | ---- | ---- | --- | ----- | --- | ------ | ------- | ------- | ----- | ----- |----- |------|-------| |
| 78 | +| EB |22.11 |31.53 |0.26 |17.65 |13.22|18.41 |14796.52 |171806.84|2311.25|2445.89|19.41 |28.89 |19.77 | |
| 79 | +| R-CNN |20.37 |44.86 |0.40 |13.81 |16.40|20.69 |22535.15 |193041.87|5029.42|5795.73|16.01 |38.35 |16.81 | |
| 80 | + |
| 81 | +##### EB detections |
| 82 | +The public detections of [EB](http://zyb.im/research/EB/) are not available on the |
| 83 | +DETRAC training set and miss some low scoring detections. The EB detections we used for the tables above and our |
| 84 | +publication are available here: |
| 85 | + |
| 86 | +* [EB Train](https://tubcloud.tu-berlin.de/s/EtC6cFEYsAU0gFQ/download) |
| 87 | +* [EB Test](https://tubcloud.tu-berlin.de/s/oKM3dYhJbMFl1dY/download) |
| 88 | + |
| 89 | +### MOT |
| 90 | +To reproduce the reported MOT16 results, use the mot16.py script: |
| 91 | + |
| 92 | +``` |
| 93 | +$ ./mot16.py -h |
| 94 | +usage: mot16.py [-h] -m SEQMAP -o RES_DIR -b BENCHMARK_DIR [-sl SIGMA_L] |
| 95 | + [-sh SIGMA_H] [-si SIGMA_IOU] [-tm T_MIN] |
| 96 | +
|
| 97 | +IOU Tracker MOT demo script. Default parameters are set to reproduce the |
| 98 | +results using the SDP detections. |
| 99 | +
|
| 100 | +optional arguments: |
| 101 | + -h, --help show this help message and exit |
| 102 | + -m SEQMAP, --seqmap SEQMAP |
| 103 | + full path to the seqmap file to evaluate |
| 104 | + -o RES_DIR, --res_dir RES_DIR |
| 105 | + path to the results directory |
| 106 | + -b BENCHMARK_DIR, --benchmark_dir BENCHMARK_DIR |
| 107 | + path to the sequence directory |
| 108 | + -sl SIGMA_L, --sigma_l SIGMA_L |
| 109 | + low detection threshold |
| 110 | + -sh SIGMA_H, --sigma_h SIGMA_H |
| 111 | + high detection threshold |
| 112 | + -si SIGMA_IOU, --sigma_iou SIGMA_IOU |
| 113 | + intersection-over-union threshold |
| 114 | + -tm T_MIN, --t_min T_MIN |
| 115 | + minimum track length |
| 116 | +``` |
| 117 | + |
| 118 | +Example: |
| 119 | +``` |
| 120 | +# SDP: |
| 121 | +./mot16.py -m ../motchallenge/seqmaps/sdp-train.txt -o ../motchallenge/res/MOT16/iou-tracker -b ../data/mot17/train |
| 122 | +
|
| 123 | +# FRCNN: |
| 124 | +./mot16.py -m ../motchallenge/seqmaps/frcnn-train.txt -o ../motchallenge/res/MOT16/iou-tracker -b ../data/mot17/train -sl 0 -sh 0.9 -si 0.3 -tm 5 |
| 125 | +``` |
| 126 | + |
| 127 | +You should obtain something like the following results for the train set: |
| 128 | + |
| 129 | +##### MOT16 Train Results |
| 130 | +| Detector | IDF1 | IDP | IDR | Rcll | Prcn | FAR | GT | MT | PT | ML | FP | FN | IDs | FM | MOTA | MOTP | MOTAL | |
| 131 | +| -------- | ---- | --- | --- | ---- | ---- | --- | --- | --- | --- | --- | --- | --- | --- | --- | ---- | ---- | ----- | |
| 132 | +|SDP |24.7 |46.2 |16.9 |65.0 |97.6 |0.34 |546 |178 |232 |136 |1796 |39348|1198 |1453 |62.3 |83.4 |63.4 | |
| 133 | +|FRCNN |21.0 |46.5 |13.6 |51.8 |97.2 |0.31 |546 |109 |261 |176 |1674 |54082|716 |810 |49.7 |88.2 |50.3 | |
| 134 | + |
| 135 | +##### MOT16 Test Results |
| 136 | +| Detector | Rcll | Prcn | FAR | GT | MT | PT | ML | FP | FN | IDs | FM | MOTA | MOTP | |
| 137 | +| -------- | ---- | ---- | --- | --- | --- | --- | --- | --- | --- | --- | --- | ---- | ---- | |
| 138 | +|SDP |61.5 |95.2 |0.96 |759 |179 |330 |250 |5702 |70278|2167 |3028 |57.1 |77.1 | |
| 139 | +|FRCNN |50.9 |92.4 |1.29 |759 |113 |381 |265 |7639 |89535| 2284|2310 |45.4 |77.5 | |
| 140 | + |
| 141 | + Please note that this evaluation already includes the new ground truth of the MOT17 release. |
| 142 | + |
| 143 | +## Contact |
| 144 | +If you have any questions or encounter problems regarding the method/code feel free to contact me |
| 145 | + |
0 commit comments