Open kruztw opened 2 years ago
Hi, as the paper said, the target models are from Apollo-5.5. But I used sha1sum to check the models in Apollo-5.5 , and did not find any models which sha1sum was same as the models you gave.
Here is the result:
root@LAPTOP-ER7AMA4V:/mnt/c/Users/kruztw/Downloads# find . | grep "deploy.caffemodel" | xargs -n 1 sha1sum | sort 0508e0bd8fe03d53093ffec31d8117cf596f1019 ./apollo-5.5.0/modules/perception/production/data/perception/lidar/models/cnnseg/velodyne128/deploy.caffemodel 3d4cfa2b4586eee92bfe14f83fb526fbe6d96ba7 ./apollo-5.5.0/modules/perception/testdata/camera/lib/lane/detector/denseline/data/denseline/deploy.caffemodel.metadata 59de55692357639fdb66e0dd67288ee3a7641192 ./apollo-5.5.0/modules/perception/testdata/camera/lib/lane/detector/denseline/data/denseline/deploy.caffemodel 727368f627d2a0ea3a0df2538afb6a51e0f03e4b ./apollo-5.5.0/modules/perception/inference/inference_test_data/denseline_parser/deploy.caffemodel.metadata 8846d41fd7a89523de94b2f19feec27cf1890d41 ./apollo-5.5.0/modules/perception/testdata/camera/lib/lane/postprocessor/denseline/data/denseline/deploy.caffemodel.metadata 8e7a0c55de7a04f5ef1976abbe73d77932d08a47 ./apollo-5.5.0/modules/perception/production/data/perception/camera/models/lane_detector/darkSCNN/deploy.caffemodel 8e7a0c55de7a04f5ef1976abbe73d77932d08a47 ./apollo-5.5.0/modules/perception/testdata/camera/lib/lane/detector/darkSCNN/data/darkSCNN/deploy.caffemodel 8e7a0c55de7a04f5ef1976abbe73d77932d08a47 ./apollo-5.5.0/modules/perception/testdata/camera/lib/lane/postprocessor/darkSCNN/data/darkSCNN/deploy.caffemodel 9689947cb4b54b2818afd38ea2c929c5338a2768 ./apollo-5.5.0/modules/perception/production/data/perception/lidar/models/cnnseg/velodyne128/deploy.caffemodel.metadata a9cdbfbd4ecd535207835c80b449046813dbba94 ./apollo-5.5.0/modules/perception/production/data/perception/lidar/models/cnnseg/velodyne16/deploy.caffemodel a9cdbfbd4ecd535207835c80b449046813dbba94 ./apollo-5.5.0/modules/perception/production/data/perception/lidar/models/cnnseg/velodyne64/deploy.caffemodel b9b485e590e3cd5ee7f9c439e0bc9b11e9aba7aa ./msf_adv_data/deploy.caffemodel c390ae582106be5b0a35062d6d334c46d964ad0f ./apollo-5.5.0/modules/perception/production/data/perception/camera/models/lane_detector/denseline/deploy.caffemodel.metadata dd990fb76feb5804136b920bac8574f6f1cb9f72 ./apollo-5.5.0/modules/perception/production/data/perception/lidar/models/cnnseg/velodyne16/deploy.caffemodel.metadata dd990fb76feb5804136b920bac8574f6f1cb9f72 ./apollo-5.5.0/modules/perception/production/data/perception/lidar/models/cnnseg/velodyne64/deploy.caffemodel.metadata ec28dbfc94b209c4848aa3ef341b5d943a535522 ./apollo-5.5.0/modules/perception/testdata/camera/lib/lane/postprocessor/denseline/data/denseline/deploy.caffemodel ee63fea7932d916b30cd50e0e642189907734e58 ./apollo-5.5.0/modules/perception/production/data/perception/camera/models/lane_detector/denseline/deploy.caffemodel
Hi @kruztw , I think I found the model they use. In the paper they state that they use both A2-L (v2.5) and A5-L (v5.5) to test the detection performance on both versions. But after a while, I think the model they include in the data.zip is from v2.0, with the same architecture and number of params, I guess they want to include a small model for convenience purpose.
Hi, as the paper said, the target models are from Apollo-5.5. But I used sha1sum to check the models in Apollo-5.5 , and did not find any models which sha1sum was same as the models you gave.
Here is the result: