![]() ![]() Sudoku Solver: Sudoku is a famous Japanese puzzle which can be found in newspapers and magazines. Object Character Recognition (OCR) model trained on MNIST dataset is used to auto detect the puzzle. Depth first search (DFS) and Binary search are used to extract words and validate through a dictionary respectively. Wrote a python program that can solve the puzzle automatically. Wordament Solver: Wordament is a puzzle game built by Microsoft, which can be found on Google and App Store. ![]() Studied state-of-art image processing and computer vision techniques in depth to obtain the best possible accuracy. Reduced overall task and human capital by 50% to 70%. Achieved a true positive rate of 95% and a false positive rate of 4% which are comparable to state-of-art results.įoreground Extraction by GrabCut: Automated extracting foreground pixels from a huge set of images (15000) using image processing techniques and visionĪlgorithms like GrabCut and GMM. Conducted object detection techniques like Faster-RCNN to detect nodules in the first phase, followed by CNN based nodule classification to reduce false positive rate. Pulmonary Nodule Detection and Lung Cancer Prediction: Designed an automatic two-phase computer aided diagnosis system that can detect nodules and assuage the arduous task of radiologists, who manually identify them. Achieved a BLEU4 score (metric to evaluate quality of generated captions) of 20.98 which trails human baseline by 0.72. Trained the model from scratch on MSCOCO dataset and investigate it by tuning minimum word frequency threshold, number of LSTM layers and type of search technique used for words sampling (Greedy or Beam). CNN LSTM Based Framework For Automated Image Captioning:Designed CNN-LSTM architecture to automatically generate sensible and human understandable descriptions for images.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |