galib360 / media_retrieval_sbir Goto Github PK

Repository for Sketch Based Image Retrieval Project Course On Media Retrieval

Description:

Folders:

1. sketchyTrain/ is the directory of sketch images used for training the model
2. sketchyTest/ is the directory of sketch images used for validation of the trained model
3. retrieved_data/ is the directory where the retrieved real images will be downloaded
4. urls/ is the directory where the retrieved urls reside
5. testing/ is the directory of images that was used to predict unseen data

Jupyter notebook:

1. Train_model_history.ipynb is the jupyter notebook file used for model training that includes all the models that we tried.
2. Retrival.ipynb is the jupyter notebook for loading the trained model, python canvas drawing, predicting and retrieving real images.

FinalTrain.py is the python script to train the model.

Model: sketchmodel125final.h5 is the saved weights from the final training.

SBIR_report.pdf is the project report.

eval.csv is the spreadsheet with accuracy measured for each of the classes against the validation dataset