OSPP Summer 2022 Logging

——————————————————————————————————————————————————–

Pre Coding Period

• After spending days and days searching for a jekyll theme for my blog, I finally found an amazing theme. So, here it is, start of my first blog.

• Last week I was little busy with my ongoing Internship & other stuff, I’m still exploring Milvus Database. No great updates but Slow & Steady wins the Race 😉️!

——————————————————————————————————————————————————–

Week 1

• Face detection and alignment are important early stages of a modern face recognition pipeline.
• Completed the procudure of Face Detection using MTCNN(Multi-Task Cascaded Convolutional Neural Networks).
• Tested the detection algorithm on various images & faces.

Face Detection Results:

Colab Notebook Link:

https://colab.research.google.com/drive/1WOPVxNc33eonhDD4gX2VGQ8zofbKm3fF?usp=sharing

——————————————————————————————————————————————————–

Week 2

• Tried & tested Face Detection using RetinaFace
• Explored Face Embeddings Extraction Methods

References:

• Research Paper- Feature Extraction
• A study of Face Embedding in Face Recognition

——————————————————————————————————————————————————–

Week 3

• Completed the implementation of Facial Embeddings Extraction Facenet & OpenCV.
• GitHub Repo [Draft]:

  https://github.com/Spnetic-5/face_rec_draft

——————————————————————————————————————————————————–

Week 4

• Loaded the FaceNet model to obtain embeddings from the images of cropped faces
• Embeddings vectors were successfully converted back to numpy arrays.
• Google Colab Notebook
• Nothing much progress this week due to placements season.

——————————————————————————————————————————————————–

Week 5

• Connected to the Milvus server & created a collection in the Database.
• Inserting vectors & Storing these embeddings on Milvus Database.
• Tried to resolve errors & warnings in the code due to Ubuntu 21.01

——————————————————————————————————————————————————–

Week 6

• Created a collection on the Milvus Database if not already present.
• Imported all the face embeddings into the created collection.
• Performed Indexing on the collection.
• Searched for the nearest neighbor of the given image using vector similarity search with Milvus.

Demo Results:

sp-lit@sp-lit:~/Desktop/OSPP/face_rec_draft/face_rec_bootcamp$ 
python celeb_finder.py test.jpg 
Running on device: cpu 💻️
Creating a collection on Milvus Database...📊️
Collection created.✅️
Loading in encoded vectors...
(51598, 512) float32 [   1    1    1 ... 3040 3040 3040]
(51598, 512) float32 [3040 3040 3040 ...  534  534  534]
(51598, 512) float32 [ 534  534  534 ... 7592 7593 7593]
(51597, 512) float32 [7593 7593 7593 ... 9999 9999 9999]
Vectors loaded in.
Indexing the data...🔢️
Indexed!✅️
Searching for the image ...🧐️
Wohoo, Similar Images/Faces found!🥳️
[8328, 5876]

Code: https://github.com/Spnetic-5/ospp_face_rec_draft/tree/main/face_rec_bootcamp

——————————————————————————————————————————————————–