Now you can Instantly Chat with Himalaya!
Skills
Positions
Portfolio Projects
Description
◦ Develop Visual Inspection (Object Detection) model to filter out faulty medical parts on conveyor line.
◦ This reduces the manual inspection cost across 10 plants for 30+ defects, using vision techniques like Yolo V3, Fast R-CNN and
Detectron2.
Research paper exploration to create an optimal DL model.
Developed object detection model and deep learning (CNN, R-CNN, FR-CNN, YOLO)
Tuned the model to optimal performance.
Used depth wise separable convolution and efficient net technique to reduce the model parameter without affecting performance.
Performed post optimization of the model by fusing its convolution, batch norm and activation layer together.
Dashboard for the live monitoring along production lines to improve SPM rate and worker efficiency reducing operational cost
Tensorflow, python and Flask was used in development phase, Docker was used in deployment phase.
Description
Studied and experimented on various DL model to create an optimal DL model for improving the diagnostics capabilities via medical
images.
◦ Increase the sharpness of edge, and increate the signal to noise ratio of the image.
◦ Developed an efficient light weight model with less hyperparameter in TensorFlow JS and python as good efficiency was achieved
due to depth wise separable convolution
◦ The Deep learning algorithm was developed to work in coordination with the image reconstruction loop of OSEM (Ordered subset
expectation Maximization)
◦ TensorFlow JS
Description
Developed CNN based deep learning model for emotion classification on video data.
◦ The model was tuned and optimized to improve the inference quality and reduce the time taken by the model.
◦ Data preprocessing – resize, color conversion, denoising, image enhancement, data augmentation using TensorFlow and python.
◦ AWS transcribe service is used for the audio sentiment classifications.
◦ TensorFlow, Tensorflow extended, and python was used in development phase and AWS cloud was used to deploy the model.
Description
◦ Helped in model selection and tuning.
◦ YOLO algorithm was used for the localization of the license plate on the image.
◦ OCR was developed separately to read the character present in the license plate.
◦ Radar was used to get the distance and velocity of the vehicle.
◦ An automated warning was generated if a vehicle crossed the red light.
◦ TensorFlow sequential and functional API was used in development phase.