$14.0 per Hour
About Parth P
Motivational Machine Learning Engineer with almost 3+ years of experience. Passionate about building models that fix problems. Relevant skills include Machine Learning, Deep Learning, Neural networks, problem-solving, programming, and creative thinking.
Tech Stack Expertise
- January 2020 - May 2023 - 3 Year
Face Recognition System
- September 2022 - May 2023 - 9 Months
This is an in-house project for the company to create an attendance system using face recognition. This system uses CCTV cameras to recognise the company employee and log when she/he comes in and out of the office. In the end, one can see the data in CSV files for each employee. For this system, a custom dataset was created by collecting the images for each employee and to achieve better accuracy transfer-learning with Yolo-v3 pre-trained weights had been used
Roles & Responsibilities: I was responsible for working on a Face recognition System for an in-house product, and collected images to create a custom dataset. Used the Yolov3 algorithm for transfer learning (trained custom data using pre-trained weights) using class labelling and data augmentation techniques to increase dataset size & got good accuracy and implemented APIs in Flask to serve data labels to the frontend for the images received.
Plate Detection & Recognition
- April 2022 - August 2022 - 5 Months
The Number Plate Detection & Recognition System enables one to identify the vehicle by capturing and detecting a moving vehicle's number plate, performing OCR and transmitting the time stamp of entry & exit to the database in real-time. This system uses computer-vision technology & has the ability to automatically identify the vehicle with the help of an image, provided by video surveillance cameras.
Roles & Responsibilities: I was responsible to build & train the models for Number Plate Detection & Recognition using Python, Yolo v7, Machine Learning, OpenCV and TensorFlow.js. Model development and model training & tuning.
3. Image Similarity
- February 2021 - March 2022 - 14 Months
This is a POC project where you compare two images and find the similarity between them. Basically, when a user passes two images in the software, the software tells them that those two images are similar or dissimilar. To find the similarity between two images the Siamese network which is CNN-based architecture and the cosine similarity algorithm to get the similarity score have been used.
Roles & Responsibilities:
I was responsible for implementing the Siamese network (CNN-based architecture) to find similarities between two images and achieved 85% accuracy on given data.
The accuracy was improved using triplet loss with binary cross-entropy and data augmentation techniques instead of the contrastive loss method. Used the Pytorch framework for data loading and implemented the Siamese network.
4. Recommendation System
- November 2019 - January 2021 - 15 Months
This is a user-based recommendation system which is using the content-based approach. In this system, when a user joins any event/meeting through the Accelevents software at that time, the system can recommend the top 5 users who have the same profile information/bio. To achieve the recommendations I used the FastText model for word-embeddings and K-Means Clustering Algorithm.
Roles & Responsibilities:
I was responsible for building the user profile-based recommendation system in Python, using a content-based approach with data cleaning and visualization.
Implementation of FastText model to get text embeddings, applied K-Means Clustering algorithm and achieved 57.45 silhouette score.
Information Technology in B.EGujarat University
- June 2016 - June 2019