PROJECTS
Image-capturing-API-for-deep-learning-dataset
A simple API using google-images-download to fetch images automatically from google search. I used google colab to run the code but can be run in any IDE. In this case images regarding yoga pose has been demostrated. To simplify the steps I have kept a copy of frozen_inference_graph.pb in this repository. Clone the repo and point out the .pb file location in the code.To download preferred images from google search simply change the name of the search files in the complete_list and make corresponding adjustment in
both download_image and detect_persons_and_save_crops functions.
The final result from a yoga database collected with this api shown on the right.
both download_image and detect_persons_and_save_crops functions.
The final result from a yoga database collected with this api shown on the right.
Creating animated images from still pictures using
CartoonGAN and Style-Transfer
I proposed a way to transform real photos into cartoon or animated style version of the images using two different networks by using neural style algorithm based on deep learning and Generative Adversarial Network (GAN) namely CartoonGAN and Style Transfer model. For the CartoonGAN model which is a generative adversarial network (GAN) framework for cartoon stylization, we implemented predefined training weight released by the author and creator of the model and we created transformer and data preprocessing method to transform the still pictures into the consecutive animated version related to the original network. We then defined Style Transfer method, a deep learning framework we used using pretrained famous VGG19 model and pretrained imagenet weights, to compare the model with the CartoonGAN.
|
|
Pet-Medicine-Search-App USING ANGULAR
Pet Medicine Search App using Angular 8, HTML5, Bootstrap and CSS to enter the description and owner of the pet and search for medicine or contact doctor
Cryptocurrency Prediction using LSTM and GRU
A simple Bidirectional LSTM and GRU model to predict cyptocurrency written in Google Colab.Look into the file Cryptocurrency_analysis.ipynb for detailed description.
We need to upload data before proceeding with the analysis. I used cryptocurrency data from Kaggle. Then we need to convert the timestamp into Date format followed by indexing and normalization of the dataset.
We need to upload data before proceeding with the analysis. I used cryptocurrency data from Kaggle. Then we need to convert the timestamp into Date format followed by indexing and normalization of the dataset.
2D-Brick-Breaker-Game in java
A simple 2D Brick Breaker Game in Java. I used Java Swing to implement the game and will work with Java FX in future to implement better version of it. The bar can be implemented using mouse but to keep the matter simple I used right and left key to move the bar. I tried to implement the game the simplest way possible according to the user input.
Image filtering pipeline using python
Filters dicom based on slice thickness and series ID and save useful information in a new csv file. This code has been developed for ImmersiveTouch Inc. which is a VR based startup specialized in healthcare based VR software and has been integrated in their pipeline.
The repository is able to do the following steps:
1. Rename the Dicom files.
2. List all the important Dicom tags into a csv file.
3. Separate the usable Dicom into another folder based on SeriesID (considering Slice thickness <=2 and pixel spacing <= 1).
4. Each Series ID is assigned to each folder within the institute folder and the folders are named as A,B,C,D and so on.
5. Finally listing all the filtered data into another CSV file.
The repository is able to do the following steps:
1. Rename the Dicom files.
2. List all the important Dicom tags into a csv file.
3. Separate the usable Dicom into another folder based on SeriesID (considering Slice thickness <=2 and pixel spacing <= 1).
4. Each Series ID is assigned to each folder within the institute folder and the folders are named as A,B,C,D and so on.
5. Finally listing all the filtered data into another CSV file.