Akshay-Dongare / Signature-Verification

Using CycleGAN, YOLOv5 and VGG16 transfer learned on custom signature datasets to check for forging of signatures.

Signature-Verification

Demo on Youtube:

Click the image below

• Or use this link: https://www.youtube.com/watch?v=im5DmQXO7RE

Steps to set up:

1. Clone the repo and cd into it
2. Downloading required models
   • Download https://huggingface.co/Akshay-Dongare/kerasVggSigFeatures.h5 and place it in SOURCE\vgg_finetuned_model
   • Download https://huggingface.co/Akshay-Dongare/bbox_regression_cnn.h5 and place it in SOURCE\vgg_finetuned_model
   • Download https://huggingface.co/Akshay-Dongare/yolov5-CustomDatasetFromTobacco800 and place it in SOURCE\yolo_files
   • Download https://huggingface.co/Akshay-Dongare/CycleGAN-Signature-Verification-Dataset and place it in SOURCE/gan_files/checkpoints/gan_signdata_kaggle
3. In SOURCE/vgg_finetuned_model/vgg_verify.py, go to line 50 and put your own model path as raw string

   Signature-Verification/SOURCE/vgg_finetuned_model/vgg_verify.py

   Line 50 in 5f6d9a1

   feature_extractor = tf.keras.models.load_model(r"C:\Users\aksha\Desktop\Git Cloned Repositories\Signature-Verification_System_using_YOLOv5-and-CycleGAN\Streamlit_App\SOURCE\vgg_finetuned_model\bbox_regression_cnn.h5")

4. Create Conda Environment (env exported from Windows 11 system)

   conda env create -f ./environment.yml
   

5. Activate Conda Environment

   conda activate signature-verification
   

6. Run Streamlit App

   streamlit run ./ui.py