Change directory to the repository root.
python main/test.py --config experiments/siamfcpp/test/got10k/siamfcpp_alexnet-got.yaml
A collection of test scripts are located under tools/test/:
EXP_NAME is the string value of key test.track.exp_name in the corresponding .yaml file.
view logs/VOT2018/<EXP_NAME>.csv
GOT-Benchmark contains testers for a series of benchmarks, including OTB, VOT, LaSOT, GOT-10k, TrackingNet.
view logs/GOT-Benchmark/report/GOT-10k/<EX015P_NAME>/performance.json
view logs/GOT-Benchmark/report/LaSOT/<EXP_NAME>/performance.json
view logs/GOT-Benchmark/report/otb2015/<EXP_NAME>/performance.json
view logs/GOT-Benchmark/report/TrackingNet/<EXP_NAME>/performance.json
-
we adopt torch2trt accelerate the speed. Because of this issue, we only transfer the feature extraction part yet.
-
please refer https://github.com/researchmm/TracKit/blob/master/lib/tutorial/install_trt.md to install torch2trt.
-
we provide trt model for siamfcpp-googlenet-vot now, you can set trt_mode True in yaml to enable it.
-
You can refer to cvt_trt.py to transfer other models to trt.
Consider changing device_num in .yaml configuration file.
tester:
names: ["GOT10kTester",]
GOT10kTester:
exp_name: *TEST_NAME
exp_save: *TEST_SAVE
device_num: 4 # change here to use four GPU
subsets: ["val"] # (val|test)