Skip to content

A project under the Computer Science division of ISTE, NITK.

Notifications You must be signed in to change notification settings

salonimathur16/Image-Cataloguer

 
 

Repository files navigation

Image Cataloguer

This project was done under the Crypt Special Interest Group of ISTE-NITK. This is the web interface which helps you interact with the image cataloguer model.

The web interface takes in an image as the input and stores it in the static/input folder. The deep learning model takes the input from the input folder and outputs the images in the static/pics folder. All the images present in the output folder are then dispalyed on the web page.

Parallelly, the Yolov4 model prints out the elements present in the images(Elements are selected from a class of 80 objects).

Also the dominant colors present in the image are displayed as tags along with the element tags.

Before running the yolo.py the files feature_list.pkl, coco.names, yolov4.cfg, yolov4.weights must be present in the folder.(same level as yolo.py)

The cataloguer dataset must be present in the static folder.(in the same level as inputs/ and pics/).

The notebook has also been included(image cataloguer3 and colors).

Commands to run the app- open the terminal and cd into this folder and type-

$env:FLASK_APP='yolo.py'
flask run

To turn on debug mode type-

$env:FLASK_APP='yolo.py'
$env:FLASK_DEBUG=1
flask run

Dataset: https://www.kaggle.com/rahulxd/image-cataloguer/

Link for COCO.Names used: https://drive.google.com/file/d/1YWagPmrNetdgydR3z7rqr_d9ABvkboQW/view?usp=sharing

Reverse Image Search Model: https://www.kaggle.com/salonimathur16/notebook86698da377

YoloV4 Model: https://www.kaggle.com/aneesh2002/yolov4-project

YoloV4 Weights: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights

YoloV4 Config: https://github.com/kiyoshiiriemon/yolov4_darknet/blob/master/cfg/yolov4.cfg

About

A project under the Computer Science division of ISTE, NITK.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 97.2%
  • Python 1.9%
  • HTML 0.9%