Deepak Karmakar Email & Phone Number

Bengaluru, Ka, In

Bangalore

Project: Realtime Auto Signage Display Setup• Designed FCN(CNN ) based regression model for identifying the Signage Display.• Implemented Auto Encoder Decoder (CNN model) for sensor data compression.• Created Vision based RSSI annotation tool using Opencv, Flask and Socket-io.• Create data pipeline using TF Record, TF Dataset, Data generator and feed to keras model.• Analysis and preprocessing of data like Cleaning, Shuffling, Augmentation, Normalization etc.• Providing ML training, mentoring and directing the team for setup, troubleshoot

Bangalore

Driver Behavior Monitoring System• Implemented CNN for driver behavior monitoring system using Keras and Tensorflow.• Implemented Classification, Regression model with custom loss function. • Porting the models to edge devices like Nvidia Jetson, Rpi, Intel Movidus NCS.Face Detection and Recognition System• Performance improvement of face detection algorithms like MTCNN, ARC Face.• Prepare dataset using hard mining and train the model end to end.• Plot training details like loss, accuracy, AUC, Confusion matrix in tensorboard.Pallet Detection and Counting System• Detect and count Pallet using Yolo object detection Algorithm.• Developed regression model for classifying the pallet into different colors.• Written algorithm for counting the box on each pallet using 5 camera.POC and live Demo• Tracking Person using Deepsort and Yolo Algorithm.• Head nodding and emotion detection for ATM antitheft alarming system.

Bangalore

Project: Device Failure Prediction and Raggedness Calculation• Designed light weight regression model for detecting pattern and correcting the skew, offset.• Training and fine tuning neural network for sensor part failure prediction in Jumbo Printer.• Writing vision algorithm like raggedness calculation and mapping algorithm for nozzle in printer.

Project: Pluck• Developing object detection algorithm like Faster RCNN, bounding box Regression, RPN.• Experimenting on custom object detection algorithms, loss functions.• Module development for locating Face, multiple object and popping the description of object.• Annotation tool development for labelling the images and save them in both xml and json format.

Project: IRPAS• Worked on Face detection and character recognition using computer vision.• Used OpenCV for edge detection, contour detection for locating the id and its information.• Integrating api services with Reliance Servers like Ifoundary and Blaze for premium calculation.

Bachelor Of Technology - Btech, Computer Science, 73

Activities and Societies: Techfest,trainer

Education record