3rd Ati-UAV Model ad Dataset

This project provides a baselie model ad evaluatio code for track1 ad track2 for CVPR23 3rd Ati-UAV workshop.

News

Test Set ca be dowloaded with followig code:

from modelscope.msdatasets import MsDataset
from modelscope.utils.costat import DowloadMode

cache_dir = '/home/ly261666/datasets'
test_set_1 = MsDataset.load('3rd_Ati-UAV', amespace='ly261666', cache_dir=cache_dir, dowload_mode=DowloadMode.FORCE_REDOWNLOAD)
prit(ext(iter(test_set_1)))

You ca also dowload dataset from the url。 Evaluate code ca be see i the sectio "Baselie Evaluatio Code for test Set" i this page.

Istallatio

coda create - ati_uav pytho=3.7
coda activate ati_uav
# pytorch >= 1.3.0
pip istall torch==1.8.1+cu102  torchvisio==0.9.1+cu102 torchaudio==0.8.1  --extra-idex-url https://dowload.pytorch.org/whl/cu102
git cloe https://github.com/ly19965/CVPR_Ati_UAV
cd CVPR_Ati_UAV
pip istall -r requiremets/tests.txt 
pip istall -r requiremets/framework.txt
pip istall -r requiremets/cv.txt 
pip istall -r requiremets/yolov5.txt

Dowload Dataset

Ru the dowload code:

from modelscope.msdatasets import MsDataset
from modelscope.utils.costat import DowloadMode


## set dataset path
cache_dir = '/home/ly261666/datasets' 

# Dowload trai set
trai_set = MsDataset.load('3rd_Ati-UAV', amespace='ly261666', split='trai', cache_dir=cache_dir, dowload_mode=DowloadMode.FORCE_REDOWNLOAD)
prit(ext(iter(trai_set)))

# Dowload validatio set
val_set = MsDataset.load('3rd_Ati-UAV', amespace='ly261666', split='validatio', cache_dir=cache_dir, dowload_mode=DowloadMode.FORCE_REDOWNLOAD)
prit(ext(iter(val_set)))

Baselie Evaluatio Code for validatio Set (Oly Support Evalutatio Code)

Evaluatio code for track1

cd CVPR_Ati_UAV
CUDA_VISIBLE_DEVICES=0  PYTHONPATH=. pytho tests/pipelies/test_ati-uav_val_track1.py
eval result: 0.125

Note:

• The validatio dataset for track1 ca be dowloaded automatically.
• Set the dataset path (lie 28) i tests/pipelies/testati-uavval_track1.py

Evaluatio code for track2

cd CVPR_Ati_UAV
CUDA_VISIBLE_DEVICES=1  PYTHONPATH=. pytho tests/pipelies/test_ati-uav_val_track2.py
eval result: 0.125

Note:

• The validatio dataset for track2 ca be dowloaded automatically.
• Set the dataset path (lie 28) i tests/pipelies/testati-uavval_track2.py

Baselie Traiig ad Evalutio Code o Validatio Set

Track 1

Ostracker Traiig code for track1

## git pull ostracker code
git cloe -b ostracker https://github.com/ly19965/CVPR_Ati_UAV

## Traiig code o Got-10k dataset
cd CVPR_Ati_UAV
PYTHONPATH=. pytho uav_scripts/trai_ostracker_got10k_l.py

## Traiig code o Ati-UAV 2023 dataset
cd CVPR_Ati_UAV
PYTHONPATH=. pytho uav_scripts/trai_ostracker_uav_l.py

Note:

• The traiig dataset (Ati-UAV-23 ad Got-10k) ca be dowloaded automatically
• Set the dataset path, cachepath ad pretraimodel path i uavscripts/traiostrackergot10kl.py ad uavscripts/traiostrackeruavl.py
• More traiig Details ca be see i OstrackerModelGot-10kL ad OstrackerModelUAVL. OstrackerModelUAVL is a pretrai model for OstrackerModelGot-10kL.

Traiig code for track1

## Traiig code o Got-10k dataset
cd CVPR_Ati_UAV
PYTHONPATH=. pytho uav_scripts/trai_siamfc_got10k.py

## Traiig code o Ati-UAV 2023 dataset
cd CVPR_Ati_UAV
PYTHONPATH=. pytho uav_scripts/trai_siamfc_uav.py

Note:

• The traiig dataset (Ati-UAV-23 ad Got-10k) ca be dowloaded automatically
• Set the dataset path, cachepath ad pretraimodel path i uavscripts/traisiamfcgot10k.py ad uavscripts/traisiamfcuav.py
• More traiig Details ca be see i SiamfcModelGot-10kS ad SiamfcModelAti-UAV-23S. SiamfcModelGot-10kS is a pretrai model for SiamfcModelAti-UAV-23S.

Evaluatio code for track1

cd CVPR_Ati_UAV
PYTHONPATH=. pytho uav_scripts/test_ati-uav_val_track1.py

Note:

• The validatio dataset (Ati-UAV-23) ca be dowloaded automatically.
• Set the modelid (lie25 damo/cvalexvideo-sigle-object-trackigsiamfc or damo/cvalexvideo-sigle-object-trackigsiamfc-uav) i uavscripts/testati-uavval_track1.py

Track 1

Traiig code for track2

cd CVPR_Ati_UAV
CUDA_VISIBLE_DEVICES=1 PYTHONPATH=. pytho uav_scripts/trai_uav_detectio.py

Note:

• The traiig dataset (Ati-UAV-23) ca be dowloaded automatically.
• Set the dataset path ad cachepath i uavscripts/traiuavdetectio.py
• More traiig Details ca be see i DamoYoloAti-UAV-23S
• The Large model will be released i DamoYoloAti-UAV-23L before 20230211

Evaluatio code for track2

cd CVPR_Ati_UAV
PYTHONPATH=. pytho uav_scripts/test_ati-uav_val_track2.py
eval result: 0.169

Note:

• The validatio dataset (Ati-UAV-23) ca be dowloaded automatically.
• Set the traker (lie24 damo/cvalexvideo-sigle-object-trackigsiamfc-uav) ad detetio modelid (lie27 damo/cvtiyasuav-detectiodamoyolo) i uavscripts/testati-uavval_track2.py
• Test your owr model:

lie 101 i uav_scripts/test_ati-uav_val_track2.py
tracker_model_path = ""
det_model_path = ""
uav_tracker.model.load_state_dcit(torch.load(tracker_model_path))
uav_detectio.model.load_state_dcit(torch.load(det_model_path))

Baselie Evaluatio Code for test Set

Note: Code is available ow, but the test set will be released i 2023.03.07.

Evaluatio code for track1

• Test SiamFC tracker with iitial target groud-truth box

cd CVPR_Ati_UAV
pytho tests/pipelies/test_ati-uav_track1.py

Note:

• The track1 dataset ca be dowloaded automatically.
• Set the dataset path (lie 28) i testati-uavtrack1.py
• You ca evaluatio other model by settig model path (lie 38) ad model etwork architecture (lie 11) i modelscope/pipelies/cv/atiuavcvpr2023_pipelie.py
• Submit track1.zip to track1: cd results/SiamFC && zip track1.zip *

You will see followig results (If IR_label.jso is available):

[001/140]  20190925_131530_1_7    IR Fixed Measure: 0.245
[002/140]  20190925_213001_1_5    IR Fixed Measure: -0.164
[003/140]  20190925_222534_1_3    IR Fixed Measure: 0.800
[004/140]  20190926_183941_1_8    IR Fixed Measure: 0.754
...
[Overall]    IR Mixed Measure:

Evaluatio code for track2

• Test SiamFC tracker ad Yolov5 without iitial target groud-truth box)
• Detect the iitial state of target usig yolov5 the track the target usig SiamFC. (You ca also use detectors ad trackers together for localizatio.)
• Submit track2.zip to track2: cd results/Yolo_SiamFC && zip track2.zip *

cd CVPR_Ati_UAV
pip istall -r yolov5/requiremets.txt
pytho tests/pipelies/test_ati-uav_track2.py

Note:

• The track2 dataset ca be dowloaded automatically.
• Set the dataset path (lie 28) i testati-uavtrack1.py
• You ca evaluatio other model by settig model path (lie 39) ad model etwork architecture (lie 12) i modelscope/pipelies/cv/atiuavcvpr2023_pipelie.py

3rd Anti-UAV Model and Dataset This project provides a baseline model and evaluation code for track1