Experienced ML Engineer

Ashish Sethi  • Peerlist

• Wrote MATLAB scripts to detect object using the web camera of laptop •  After detection of object it draws the line on the 2D plane of MATLAB as the object moves and   trace pattern can be save as image file. •This project uses the Image Processing tool box and GUI of the MATLAB.

Developed a Real Time Object T

Using Genetic Algorithm find the best architecture design of CNN’s network for classification. Developed two encoding techniques in genetic algorithm to find best architecture design.

Genetic CNN

• Developed a bot which runs on the MATLAB GUI buttons • Uses MATLAB GUI and ATmega 328 (Arduino UNO) as microcontroller in bot. •It uses the Serial Communication between the microcontroller and MATLAB. •I have used Two servos motor to move in 4 axis and laser light to target the water canon at fire. •On the screen of GUI, we can also see the angle at which both the servos are present at that time. •It can throw water up to distance of more than one metre which we can target using laser light    which is also embedded on bot

MATLAB GUI based Water cannon 

Developed the natural language based object retrieval method. Localize a target object in given image based on natural language query. and draw bounding box on it.

Natural Language Object Retrie

Segmentation of infant’s brain is difficult as the intensity ranges of voxels in White Matter (WM), Grey Matter (GM) and Cerebrospinal Fluid (CSF) areas are mainly overlapping. We are using residual based 3D-UNet architecture to segment the MRI images into mainly 3 regions.

NeoNet: A Deep CNN for 6-month

• We developed two robots one is ECO Bot and other is Hybrid Bot • I managed Electronics part of ECO Bot.

Participated in ABU ROBOCON

Using reinforcement learning algorithm policy for unlabelled data selection in semi-supervised learning algorithm. For semi-supervised learning, We are using co-training algorithm to train image classifier with labelled and unlabelled data.

Reinforced Co -Training

Detect change in urbanization, using satellite images over two time period. Developed Siamese U-Net based segmentation network to detect the changes in two images.

Urbanization Detection

• Wrote MATLAB Code using Image processing tool box and GUI toolboxes. •In this project camera detects the three different objects which are selected by user and this code find the distance between them and it also draws the lines between them but if objects are less than three is will not draw lines between them. •And this program also shows in which quadrant the object is present.