|
| 1 | + |
| 2 | +Zipped NN models |
| 3 | +================ |
| 4 | + |
| 5 | +The folder consists of the list of ZIP archivated models. They can be |
| 6 | +directly downloaded, unpacked on local PC. |
| 7 | + |
| 8 | +Donwnoad and unpack utility |
| 9 | +=========================== |
| 10 | +Instead of manual downloading/unpacking you can use special utility |
| 11 | +`donwnoad_unpack_cnn_models.py` |
| 12 | + |
| 13 | +## Usage Instructions |
| 14 | +1. Download `donwnoad_unpack_cnn_models.py` and `model_list.txt` |
| 15 | +2. Edit `model_list.txt` |
| 16 | + |
| 17 | + Downloading and unpacking of all models can take formidable time |
| 18 | + expecially in case of slow Internet. Do reduce time customers reduce |
| 19 | + number of models. Make copy of `model_list.txt` to |
| 20 | + `model_list_shorten.txt` and remove models you don't want to use now. |
| 21 | + You can download other models lately |
| 22 | + |
| 23 | +3. Run the script |
| 24 | + |
| 25 | + `python donwnoad_unpack_cnn_models.py --model_list |
| 26 | + model_list_shorten.txt` |
| 27 | + It will download and unpack models in path that is defined as |
| 28 | + %EV_CNNMODELS_HOME%\caffe_models, where %EV_CNNMODELS_HOME% - |
| 29 | + enviroment variable |
| 30 | + You can select your path instead |
| 31 | + `python donwnoad_unpack_cnn_models.py --model_list |
| 32 | + model_list_shorten.txt --model_path <your_path>` |
| 33 | + |
| 34 | +4. Add new models |
| 35 | + |
| 36 | + You can add new models in `model_list_shorten.txt` and re-start |
| 37 | + `python donwnoad_unpack_cnn_models.py --model_list |
| 38 | + model_list_shorten.txt` |
| 39 | + It will download/unpack only missed parts |
| 40 | + |
| 41 | +>**Note:** |
| 42 | +The script does not compare content of ZIP files and outputs. If you |
| 43 | +know that there are some changes in models you should remove |
| 44 | +`%EV_CNNMODELS_HOME%\caffe_models\<model_name>` folder and re-run script |
| 45 | +It will re-downloaded and unpack the model with the latest content |
| 46 | + |
0 commit comments