vutrungnghia99/benchmark_baselines repository overview

• conda environments:

• extract_matches, baselines: env for baselines source code (https://github.com/vcg-uvic/image-matching-benchmark-baselines⁠)
• sfm: env for benchmark (https://github.com/vcg-uvic/image-matching-benchmark⁠).

• Run baselines: mount source code and data (test, val: images only)

• sudo docker run --mount src=/media/vutrungnghia/New\ Volume/PROJECT-2/baselines,target=/home/workspace,type=bind -it vutrungnghia99/benchmark_baselines:latest
  • NOTE: ls -al /media/vutrungnghia/New Volume/PROJECT-2/baselines
    • image-matching-benchmark-baselines
    • imw-2020
• cd /home/workspace/image-matching-benchmark-baselines
• conda activate extract_patches
• python detect_sift_keypoints_and_extract_patches.py --subset=test
• conda activate baseline
• python extract_descriptors_hardnet.py --subset=test

• Run benchmark: mount source code and data (val: all sequences)

• sudo docker run --mount src=/media/vutrungnghia/New\ Volume/PROJECT-2/benchmark,target=/home/workspace,type=bind -it vutrungnghia99/benchmark_baselines:latest
  • NOTE: ls -al /media/vutrungnghia/New\ Volume/PROJECT-2/benchmark
    • image-matching-benchmark
    • data
• cd /home/workspace/image-matching-benchmark
• conda activate sfm
• python run.py --json_method=example/configs/example-test.json --subset=val --run_mode=interactive
  • NOTE: BecauseDEGENSAC and GC-RANSAC haven't installed yet => change method to MAGSAC in example/configs/example-test.json as bellow:

"geom": { "method": "cmp-magsac-f", "threshold": 0.5, "confidence": 0.999999, "max_iter": 100000, // "error_type": "sampson", // "degeneracy_check": true, },