Skip to content
BoxMOT CI

follow tracker name convention #1924

Sign in to view logs

follow tracker name convention

follow tracker name convention #1924

Workflow file for this run

.github/workflows/ci.yml at 57c6f88
# name of the workflow, what it is doing (optional)
name: BoxMOT CI
# events that trigger the workflow (required)
on:
push:
# pushes to the following branches
branches:
- master
pull_request:
# pull request where master is target
branches:
- master
jobs:
tracking-methods:
runs-on: ${{ matrix.os }}
outputs:
status: ${{ job.status }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest, macos-14] # skip windows-latest for
python-version: ['3.11']
# leads to too many workflow which ends up queued
# tracking-method: [hybridsort, botsort, ocsort, bytetrack]
# Timeout: https://stackoverflow.com/a/59076067/4521646
timeout-minutes: 50
steps:
- uses: actions/checkout@v4 # Check out the repository
- name: Set up Python
uses: actions/setup-python@v5 # Prepare environment with python 3.9
with:
python-version: ${{ matrix.python-version }}
cache: 'pip' # caching pip dependencies
- name: Install requirements
shell: bash # for Windows compatibility
run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with yolo
- name: Generate detections and embeddings
run: |
python tracking/val.py generate_dets_embs --source ./assets/MOT17-mini/train --yolo-model yolov10n.pt --reid-model osnet_x0_25_msmt17.pt --imgsz 320
- name: Run tracking method
run: |
for tracker in $TRACKERS; do
python tracking/val.py generate_mot_results --yolo-model yolov10n.pt --reid-model osnet_x0_25_msmt17.pt --imgsz 320 --tracking-method $tracker
done
evolution:
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest] # skip windows-latest for
python-version: ['3.9', '3.11']
outputs:
status: ${{ job.status }}
# Timeout: https://stackoverflow.com/a/59076067/4521646
timeout-minutes: 50
steps:
- uses: actions/checkout@v4 # Check out the repository
- name: Set up Python
uses: actions/setup-python@v5 # Prepare environment with python 3.9
with:
python-version: ${{ matrix.python-version }}
cache: 'pip' # caching pip dependencies
- name: Install requirements
shell: bash # for Windows compatibility
run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with yolo,evolve
- name: Evolve set of parameters for selected tracking method
run: |
# reuse first set of generated det and prod
python tracking/evolve.py --benchmark MOT17-mini --yolo-model yolov8n.pt --reid-model osnet_x0_25_msmt17.pt --n-trials 3 --tracking-method strongsort --source ./assets/MOT17-mini/train --ci
mot-metrics-benchmark:
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest] # skip windows-latest for
python-version: ['3.9', '3.11']
outputs:
status: ${{ job.status }}
env:
TRACKERS: "ocsort bytetrack botsort hybridsort deepocsort imprassoc strongsort"
# Timeout: https://stackoverflow.com/a/59076067/4521646
timeout-minutes: 50
steps:
- uses: actions/checkout@v4 # Check out the repository
- name: Set up Python
uses: actions/setup-python@v5 # Prepare environment with python 3.9
with:
python-version: ${{ matrix.python-version }}
cache: 'pip' # caching pip dependencies
- name: Install requirements
shell: bash # for Windows compatibility
run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
sudo apt-get install -y jq
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install
- name: Generate detections and embeddings
run: |
python tracking/val.py generate_dets_embs --source ./assets/MOT17-mini/train --yolo-model yolov10n.pt --reid-model osnet_x0_25_msmt17.pt --imgsz 320 --classes 0
- name: Evaluation and Summarize Results
shell: bash
run: |
echo "Format,Status❔,HOTA,MOTA,IDF1" > results.csv
for tracker in $TRACKERS; do
if python3 tracking/val.py --benchmark MOT17-mini --yolo-model yolov8n.pt --reid-model osnet_x0_25_msmt17.pt --tracking-method $tracker --verbose --source ./assets/MOT17-mini/train --ci; then
STATUS="✅"
else
STATUS="❌"
fi
cat ${tracker}_output.json
# Extract the relevant parts of the JSON
HOTA=$(jq -r '.HOTA' ${tracker}_output.json)
MOTA=$(jq -r '.MOTA' ${tracker}_output.json)
IDF1=$(jq -r '.IDF1' ${tracker}_output.json)
TRACKER_NAME=$(echo $tracker | awk '{print toupper(substr($0,1,1)) tolower(substr($0,2))}')
echo "$TRACKER_NAME,$STATUS,$HOTA,$MOTA,$IDF1" >> results.csv
done
# Sort the results by HOTA in descending order
(head -n 1 results.csv && tail -n +2 results.csv | sort -t, -k3 -nr) > sorted_results.csv
# Create a pretty table from the sorted_results.csv file
column -s, -t sorted_results.csv > pretty_results.txt
- name: Show Results
shell: bash
run: |
cat pretty_results.txt
tracking-with-pose:
runs-on: ubuntu-latest
outputs:
status: ${{ job.status }}
steps:
- uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: '3.11'
- run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with yolo
- name: Test tracking with pose models
env:
IMG: ./assets/MOT17-mini/train/MOT17-02-FRCNN/img1/000001.jpg
run: python tracking/track.py --yolo-model weights/yolov8n-pose.pt --source $IMG --imgsz 320
tracking-with-yolos:
runs-on: ubuntu-latest
outputs:
status: ${{ job.status }}
steps:
- uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: '3.11'
- run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with yolo
- name: Test tracking with pose models
env:
IMG: ./assets/MOT17-mini/train/MOT17-02-FRCNN/img1/000001.jpg
run: |
python tracking/track.py --yolo-model yolov10n.pt --source $IMG --imgsz 320
# python tracking/track.py --yolo-model yolox_n.pt --source $IMG --imgsz 320
# python tracking/track.py --yolo-model yolo_nas_s.pt --source $IMG --imgsz 320
tracking-with-seg:
runs-on: ubuntu-latest
outputs:
status: ${{ job.status }}
steps:
- id: set_result
run: echo "::set-output name=result::success"
- uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: '3.11'
- name: Install dependencies
run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with yolo
- name: Test tracking with seg models
env:
IMG: ./assets/MOT17-mini/train/MOT17-02-FRCNN/img1/000001.jpg
run: python tracking/track.py --tracking-method deepocsort --yolo-model yolov8n-seg.pt --source $IMG
export-reid-models:
runs-on: ubuntu-latest
outputs:
status: ${{ job.status }}
steps:
- id: set_result
run: echo "::set-output name=result::success"
- uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: '3.11'
- name: Install dependencies
run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with export
# needed for TFLite export
sudo apt-get install flatbuffers-compiler
wget https://github.com/PINTO0309/onnx2tf/releases/download/1.16.31/flatc.tar.gz
tar -zxvf flatc.tar.gz
sudo chmod +x flatc
sudo mv flatc /usr/bin/
- name: Test export models
run: |
python boxmot/appearance/reid_export.py --include torchscript onnx openvino tflite --device cpu --batch-size 3 --dynamic
- name: Check export TFLite models
run: |
ls /home/runner/work/boxmot/boxmot/tracking/weights/osnet_x0_25_msmt17_saved_model/
- name: Test inference on exported models
env:
IMG: ./assets/MOT17-mini/train/MOT17-02-FRCNN/img1/000001.jpg
run: |
python tracking/track.py --reid-model tracking/weights/osnet_x0_25_msmt17.torchscript --source $IMG --imgsz 320
python tracking/track.py --reid-model tracking/weights/osnet_x0_25_msmt17.onnx --source $IMG --imgsz 320
python tracking/track.py --reid-model tracking/weights/osnet_x0_25_msmt17_openvino_model --source $IMG --imgsz 320
python tracking/track.py --reid-model tracking/weights/osnet_x0_25_msmt17_saved_model/osnet_x0_25_msmt17_float32.tflite --source $IMG --imgsz 320
tests:
runs-on: ubuntu-latest
outputs:
status: ${{ job.status }}
steps:
- uses: actions/checkout@v4
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: '3.11'
- run: |
if [[ "$OSTYPE" == "darwin"* ]]; then
# macOS
sed -i '' 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
elif [[ "$OSTYPE" == "linux-gnu"* ]]; then
# Linux
sed -i 's/source="torch_cuda121"/source="torchcpu"/g' pyproject.toml
fi
python -m pip install --upgrade pip setuptools wheel poetry
poetry config virtualenvs.create false
poetry lock --no-update
poetry install --with test
- name: Test export models
run: python boxmot/appearance/reid_export.py --include torchscript onnx openvino --device cpu --batch-size 3 --dynamic
- name: Pytest tests # after tracking options as this does not download models
env:
# directory of PyPi package to be tested
PACKAGE_DIR: boxmot
# minimum acceptable test coverage
COVERAGE_FAIL_UNDER: 25
shell: bash # for Windows compatibility
run: |
pytest --cov=$PACKAGE_DIR --cov-report=html -v tests
coverage report --fail-under=$COVERAGE_FAIL_UNDER
# test-gpu:
# runs-on: gpu-latest
# outputs:
# status: ${{ job.status }}
# steps:
# - uses: actions/checkout@v4
# - name: Set up Python
# uses: actions/setup-python@v5
# with:
# python-version: '3.11'
# - run: |
# python -m pip install --upgrade pip setuptools wheel poetry
# poetry config virtualenvs.create false
# poetry install --with test
# - name: Pytest tests # after tracking options as this does not download models
# env:
# # directory of PyPi package to be tested
# PACKAGE_DIR: boxmot
# # minimum acceptable test coverage
# COVERAGE_FAIL_UNDER: 25
# shell: bash # for Windows compatibility
# run: |
# pytest --cov=$PACKAGE_DIR --cov-report=html -v tests/test_cuda.py
check-failures:
needs:
- tracking-methods
- mot-metrics-benchmark
- evolution
- export-reid-models
- tests
- tracking-with-pose
- tracking-with-seg
- tracking-with-yolos
if: always() # This ensures the job runs regardless of previous job failures
runs-on: ubuntu-latest
steps:
- name: Prepare environment variables
run: |
echo "tracking-methods_STATUS=${{ needs.tracking-methods.result }}" >> $GITHUB_ENV
echo "mot-metrics_STATUS=${{ needs.mot-metrics.result }}" >> $GITHUB_ENV
echo "evolution_STATUS=${{ needs.evolution.result }}" >> $GITHUB_ENV
echo "export-reid-models_STATUS=${{ needs.export-reid-modelsn.result }}" >> $GITHUB_ENV
echo "tests_STATUS=${{ needs.tests.result }}" >> $GITHUB_ENV
echo "tracking-with-pose_STATUS=${{ needs.tracking-with-pose.result }}" >> $GITHUB_ENV
echo "tracking-with-seg_STATUS=${{ needs.tracking-with-seg.result }}" >> $GITHUB_ENV
echo "tracking-with-yolos_STATUS=${{ needs.tracking-with-yolos.result }}" >> $GITHUB_ENV
- name: Check for failures and create summary
run: |
summary=""
failed=false
# Print all environment variables, grep for those ending with _STATUS, then loop
for var in $(printenv | grep '_STATUS$'); do
job_status="${var##*=}" # Extract the status part
job_name="${var%%=*}" # Extract the job name part
if [[ "$job_status" != "success" ]]; then
summary+="$job_name failed with status: $job_status\n"
failed=true
fi
done
if [[ "$failed" = false ]]; then
summary="All jobs succeeded."
fi
echo "Summary: $summary"