Drone Eye

Target

• Task 2: Object Detection in Videos in VisDrone

Trainig Data Anaylsis

The Traning data is same with 2018 version.

Result Format

Both the ground truth annotations and the submission of results on test data have the same format for object detection in videos. That is, each text file stores the detection results of the corresponding video clip, with each line containing an object instance in the video frame. The format of each line is as follows:

<frame_index>,<target_id>,<bbox_left>,<bbox_top>,<bbox_width>,<bbox_height>,<score>,<object_category>,<truncation>,<occlusion>

Please find the example format of the submission of results for object detection in videos here (BaiduYun|Google Drive)

Details Link

Sample Result (20190613)

• Green box : Ground Truth
• Red box : our result

Sample mAP Result (20190711) for a val sequence

START Calculating mAP of sequence uav0000137_00458_v (without split)

8.83% = bicycle AP

63.49% = car AP

22.67% = motor AP

29.45% = pedestrian AP

33.63% = people AP

4.38% = van AP

mAP = 27.07%

START Calculating mAP of sequence uav0000137_00458_v (with split)

26.48% = bicycle AP

50.55% = car AP

28.09% = motor AP

30.07% = pedestrian AP

43.15% = people AP

19.91% = van AP

mAP = 33.04%

How to Use It

Set up the test environment

Before the setting the environment, please use virtualenv and then install necessary libraries like below

pip install -r requirement

Submit the result

To submit the result, please use test.py like below.

$ python test.py