Soham Basu
Deep Learning, AutoML and Computer Vision Enthusiast
Location
Freiburg im Breisgau, Germany
soham.basu07@gmail.com
Deep Learning, AutoML and Computer Vision Enthusiast
Location
Freiburg im Breisgau, Germany
soham.basu07@gmail.com
Graduate student at the University of Freiburg, working towards a Masters degree in Embedded Systems Engineering with a
specialization in Artificial Intelligence.
I have a strong passion for AutoML and Deep Learning, especially Machine Vision. Still majorly in the learning phase and currently working on
Multi-objective, Multi-fidelity Hyperparameter Optimization.
Since October 2022
Freiburg im Breisgau, Germany
Master of Science - Embedded Systems Engineering
Specialization: Artificial Intelligence
GPA: 1.7 (Best possible grade: 1.0, Passing grade: 4.0)
Master Thesis: Multi-objective Hyperparameter Optimization with Expert Priors
2016 - 2020
Kolkata, India
Bachelor of Technology - Electronics and Communicaton Engineering
GPA: 7.75
Bachelor Project: Particle Swarm Optimization
2002 - 2016
Kolkata, India
High School - ISC (12) - Science
Avg: 95.5%
Since October 2023
Freiburg im Breisgau, Germany
August 2020 - September 2022
Kolkata, India
- Bone Fracture Detection and Classification (Since 2022)
Assisting in the development of deep learning-based algorithms to
detect & classify bone joints and fractures from X-Ray and CT scans.
- Diabetic Retinopathy Feature Extraction – Blood Vessels, Optic Disc and Exudates (2020 - 2021)
Extracted optic disc, blood vessels
and exudates from DR affected fundus images, using morphological transformations, k-means clustering,
edge detection and contouring. Blood Vessel Segmentation achieved state-of-the-art accuracy of 95.93% on the DRIVE dataset.
- Binary Diagnosis of Diabetic Retinopathy using Deep CNN (2020 – 2021)
Built a Deep Convolutional
Neural Network with 14 Conv2D layers and 2 Full Connected Layers (using TensorFlow Keras). Trained it on the publicly available IDRiD Dataset for Binary
Classification of Diabetic Retinopathy.
Test Accuracy: 75.73%
November 2020 - December 2021
Kolkata, India
Focused on research in the Computer Vision domain – working on the implementation of research papers in Image Denoising and Super-resolution.
April 2019 - October 2020
Kolkata, India
- Creation of online marketing strategies, handling Facebook Ads Manager.
- Frontend Web Development,
Server management and Backend integrations using PHP, MySQL on Apache HTTP Server.
June 2019
Kolkata, India
-Learned about the segments of Powergrid, their functioning, working of POWERTEL'S Region Control Centers (RTACC)
and transmission of data through OPGW wires.
- Familiarized with telecom technologies and services viz. OPGW,
CDNs, VPN, Peering, OTT, etc.
- Researched on the latest ventures - 100G fibers and MPLS networks.
April 2017 - May 2018
Kolkata, India
- Created basic, professional chatbots using Dialogflow for automated end-user interactions.
- Familiarized with
other NLP/NLU platforms like: IBM Watson, Microsoft Bot Framework, etc.
- Documented various modules developed
and module testing.
Since August 2020
Kolkata, India
- Bone Fracture Detection and Classification (Since 2022)
Assisting in the development of deep learning-based algorithms to
detect & classify bone joints and fractures from X-Ray and CT scans.
- Diabetic Retinopathy Feature Extraction – Blood Vessels, Optic Disc and Exudates (2020 - 2021)
Extracted optic disc, blood vessels
and exudates from DR affected fundus images, using morphological transformations, k-means clustering,
edge detection and contouring. Blood Vessel Segmentation achieved state-of-the-art accuracy of 95.93% on the DRIVE dataset.
- Binary Diagnosis of Diabetic Retinopathy using Deep CNN (2020 – 2021)
Built a Deep Convolutional
Neural Network with 14 Conv2D layers and 2 Full Connected Layers (using TensorFlow Keras). Trained it on the publicly available IDRiD Dataset for Binary
Classification of Diabetic Retinopathy.
Test Accuracy: 75.73%
First Online: 8th April, 2023
){kind=icon}
Soumalya Bose, Soham Basu, Indranil Bera, Sambit Mallick, Snigdha Paul, Saumodip Das, Swarnendu Sil, Swarnava Ghosh and Anindya Sen
Proceedings of Computational Vision and Bio-Inspired Computing pp 637-658
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 1439)
This paper addresses the medical imaging problem of joint detection in the upper limbs, viz. elbow, shoulder, wrist and finger joints. Localization of joints from X-ray and computerized tomography (CT) scans is
an essential step for the assessment of various bone-related medical conditions like osteoarthritis, rheumatoid arthritis, and can even be used for automated bone fracture detection. Automated joint localization also detects the corresponding bones and can serve as input
to deep learning-based models used for the computerized diagnosis of the aforementioned medical disorders. This increases the accuracy of prediction and aids the radiologists with analyzing the scans, which is quite a complex and exhausting task. This paper provides a detailed
comparative study between diverse deep learning (DL) models—YOLOv3, YOLOv7, EfficientDet and CenterNet in multiple bone joint detections in the upper limbs of the human body. The research analyzes the performance of different DL models, mathematically, graphically and visually.
These models are trained and tested on a portion of the openly available musculoskeletal radiographs (MURA) dataset. The study found that the best mean average precision (mAP0.5:0.95) values of YOLOv3, YOLOv7, EfficientDet and CenterNet are 35.3, 48.3, 46.5 and 45.9, respectively.
Besides, it has been found that YOLOv7 performed the best for accurately predicting the bounding boxes, while YOLOv3 performed the worst in the visual analysis test.
See publication
First Online: 11th July, 2021
Soham Basu, Sayantan Mukherjee, Ankit Bhattacharya and Anindya Sen
Proceedings of Research and Applications in Artificial Intelligence pp 173-184
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 1355)
Diabetic Retinopathy (DR) is a complication of long-standing, unchecked diabetes, and one of the leading causes of blindness in the world. This paper focuses on improved and robust methods to extract some of the features
of DR, viz., Blood Vessels and Exudates. Blood vessels are segmented using multiple morphological and thresholding operations. For the segmentation of exudates, k-means clustering and contour detection on the original
images are used. Extensive noise reduction is performed to remove false positives from the vessel segmentation algorithm’s results. The localization of optic disc using k-means clustering and template matching is also
performed. Lastly, this paper presents a Deep Convolutional Neural Network (DCNN) model with 14 Convolutional Layers and 2 Fully Connected Layers, for the automatic, binary diagnosis of DR. The vessel segmentation,
optic disc localization and DCNN achieve accuracies of 95.93%, 98.77%, and 75.73%, respectively.
See publication
2020
DeepLearning.AI, Coursera
Completed courses on Introduction to TensorFlow, CNNs, NLP, LSTMS, Sequences and Time Series in TensorFlow.
Certificate Link
2020
DeepLearning.AI, Coursera
Used CNNs to classify Real-World images, explored overfitting prevention strategies viz. Augmentation, Regularization and Dropouts, implemented transfer learning and extracted learned features from models.
Certificate Link
2020
DeepLearning.AI, Coursera
Learned and applied the concepts of Logistic Regression and Deep Neural Networks using Python and Numpy.
Certificate Link
2020
DeepLearning.AI, Coursera
Learned NLP basics – sequences, embeddings, etc. Built NLP systems and applied RNNs, LSTMs in TensorFlow.
Certificate Link
2020
DeepLearning.AI, Coursera
Solved Time Series and forecasting problems in TensorFlow, applied CNNs and RNNs in Real-World problems.
Certificate Link
2020
DeepLearning.AI, Coursera
Implemented the concepts of regression and classification, built Deep Convolution Neural Networks using TensorFlow 2.0. Implemented Image Data Generator and solved Real-World Image classification problems.
Certificate Link
2020
Google Cloud
Learned API hierarchy, lazy evaluation, implemented Estimator API and distributed training on Google Cloud.
Certificate Link
2018
IIT Guwahati
Gained hands-on experience on basic IOT using NodeMCU.
Hyperparameter Optimization - Bayesian Optimization, Multi-objective and Multi-fidelity HPO Optimization with Priors
Deep Learning Optimizers, Real world computer vision problems viz. Image Denoising, Super Resolution, 2D to 3D Translation, Semantic Segmentation, Estimaton.
2017 & 2020
Kolkata, India
Participated in IBM Master the Mainframe in 2017 and 2020 and reached Level 2 in 2020.
October 2016
Kolkata, India
Received an Award for the Excellence in English in High School, which is given to those with the best curricular performance in English in the12th standard and the ISC Board Examinations.
May 2016
Kolkata, India
Ranked 2nd in the ISC 2016 batch in High School and 1st among the Science students.
Art is something that has always come naturally to me. I have been painting since I was 5 years old and have recently, kind of, reached a graduation level in Art. In 2019 I received a Diploma in Fine Arts from the Ministry
of Education in India (formerly Ministry of Human Resources Development - MHRD).
I won't lie, art makes me happy, really happy - every stroke of the brush or shade of charcoal.
Check out my recent art projects
here.
I love taking photographs - not of everything, though. I mostly take photos of wildlife, birds and at times, really unique architecture. Certain weird compositions amuse me too. Photography isn't something I'm an expert in,
but something I whole-heartedly enjoy.
Check out some of my best photographs here.
I write stories (short ones), I rant, I post musings, tried my hand at poems and I document trips. My writing and English have won me a couple of awards in high school too. They were fun at the time. Competitions are irrelevant
now. I only write when I feel like. The styles or content are a direct reflection of my mood at the time of writing - although, not in the way you might expect.
Don't quite understand what I'm talking about? You'll find
most of my writings on my blog here. I'll let you be the judge of my work.
Just like art and writing, I've spent hours on this particular hobby. Heck! I've even made some money out of it. It's fun, it takes my mind off things, is another way to vent my creative instincts and I get to learn new things
everytime I develop a website.
I made a couple websites in the past few years. Check them out here.