mrkieumy / task-conditioned Goto Github PK

This source code implements our ECCV paper "task-conditioned domain adaptation for pedestrian detection in thermal imagery".

Home Page: https://github.com/mrkieumy/task-conditioned

License: MIT License

Python 85.81% Makefile 0.27% C 7.30% C++ 0.32% Cuda 6.30%

computer-vision object-detection pedestrian-detection thermal-imaging yolov3-kaist domain-adaptation transfer-learning fine-tuning cnn detector

task-conditioned's People

Contributors

Stargazers

Watchers

Forkers

hello-trouble aust-hansen ciodar aminemarnissi wassboub cqyan521 scwuchung weichiheng liting1018

task-conditioned's Issues

How to train it on FLIR

hey， i want to know that how to train the model on FLIR. The annotations file is *.xml, what should i do to change *.xml to *.txt?

About confusing comparison in experiments

In 4.1, you guys say you use the improved training annotationsfrom [22] and test annotations from [25]. so the results you obtained is evaluated on the improved test annotations , but in Table 3, methods above MSDS-RCNN like IATDNN+IAMSS with 26.37MR are reported on the original test annotations and it has a 14.95MR performance on the improved test annotations as you used, but methods like MSDS-RCNN and your results are reported on the improved test annotations,
this also happens in other comparison experiments like single-modality detectors in table 2, which is quite unfair and confusing

[22]Li, C., Song, D., Tong, R., Tang, M.: Multispectral pedestrian detection via simul-taneous detection and segmentation. In: Proc. of British Machine Vision Confer-ence (BMVC) (2018)
[25]Liu, J., Zhang, S., Wang, S., Metaxas, D.N.: Multispectral deep neural networksfor pedestrian detection. arXiv preprint arXiv:1611.02644 (2016)

Result of 'Ours TC Visible' in Table 2

Hi,

Thank you very much for this very interesting work!

I have a question about the production of 'Ours TC Visible' shown in Table 2. Are you using 'cfg/yolov3_kaist.cfg' and 'kaist_sisible_detector.weights' to produce the results? Could you please help to clarify?

Bests,
Xingchen

where is the SOTA results?

I want to plot the curve of miss rate. However, the necessary files of other methods are missing. Thank you.

detections_all['MSDS'] = parse_detections('det_coco','results/SOTA/MSDS.json')

detections_all['MSDS_sanitized'] = parse_detections('det_coco','results/SOTA/MSDS_sanitized.json')

detections_all['IAF'] = parse_detections('det_coco','results/SOTA/IAF.json')

detections_all['Early fusion'] = parse_detections('det_coco','results/SOTA/early_fusion.json')

detections_all['Late fusion'] = parse_detections('det_coco','results/SOTA/late_fusion.json')

detections_all['RCNN thermal'] = parse_detections('det_coco','results/SOTA/KAIST_thermal.json')

detections_all['RCNN rgb'] = parse_detections('det_coco','results/SOTA/KAIST_rgb.json')

detections_all['YOLO TLV'] = parse_detections('det_coco','results/SOTA/yolov2_VLT.json')

detections_all['Bottom-up'] = parse_detections('det_coco','results/SOTA/bottom_up.json')

detections_all['Ours visible'] = parse_detections('det_coco','results/ours/ours_visible.json')

detections_all['Ours TC Det'] = parse_detections('det_coco','results/ours/tc_det.json')

adaptation from visible to thermal

hi,if I want to adapt visible to thermal ,what should the parameter "adaptation" be ? what the number is in your experiments?hope for yout reply,thanks.