visionhub-ai

visionhub-ai is a Django-based application that demonstrates a powerful pipeline combining YOLOv11n-seg (instance segmentation) with bird classification and face recognition features.

Pipeline Overview

The application uses the yolo11n-seg.pt model, trained on the COCO dataset, to perform instance segmentation this includes detecting objects and generating pixel-level segmentation masks.

From the 80 COCO classes, we focus on two: bird and person. Based on which is detected, the pipeline branches into two modules:

---

Bird Detection and Classification

• When a bird is detected by yolo11n-seg, a "View Bird" button appears.
• Clicking the button navigates to the Bird Classification page.
• On that page, clicking "Classify Bird" runs the bird_resnet_classifier.pt model.
• This model is trained to classify bird species using a custom dataset.
• You can explore the training notebook here:
  Bird Classification using PyTorch (Kaggle)
• The bird images used for testing were taken from Pixabay.

Demo Video:
Watch Bird Classification Demo

---

Face Detection and Recognition

• When a person is detected by YOLOv11n-seg, a "View Face" button appears.
• Clicking it redirects the user to the Face Recognition page.
• On this page, clicking "Recognize Face" runs the face recognition logic.
• It uses the face_recognition library, which provides a pretrained model based on dlib's ResNet.
• If the detected face matches any in the pre-stored images folder, it’s labeled with a bounding box and name.

No training was done for face recognition we use pretrained embeddings to compare known and unknown faces.

Dataset used:
Celebrity Face Dataset (Kaggle) This dataset was originally created by the author listed above. For this project, it has been slightly modified, and the face images used for testing were also taken from it.

Demo Video:
Watch Face Recognition Demo

---

Technology Stack

Frontend:

• HTML
• CSS
• JavaScript

Backend:

• Django (Python)

Models and Libraries Used:

• YOLOv11n-seg (for instance segmentation)
• bird_resnet_classifier.pt (custom ResNet model for bird classification)
• face_recognition (for face recognition using pretrained face embeddings)

---

Summary

visionhub-ai demonstrates how to combine instance segmentation with classification and recognition features in a modular and interactive Django web application.