own_repos

Ideas developed or integrated with other publicly available projects, this repository is detailed as follows:

1. machine_and_deep_learning: Experiments conducted on machine and deep learning algorithms, where 3 main frameworks were used: Caffe, Tensorflow, pytorch

   a. pytorch_models/Diffusion/: own pytorch implementation of models such as Denoising Diffusion Probabilistic Model (DDPM) with class conditioning and multi-gpu support.

   b. FarePredictor: Exprimenting with Machine Learning models for predicting Taxi ride fares.

   c. Tensorflow_models/Autoencoders: Own tensroflow implementation of Denoising AutoEncoders (DAE), and AutoEncoders (AE)

   d. Tensorflow_models/DHM_segmentation_detection: An attempt to replicate the results from Deep Hierarchical Models for Joint Object Detection, which could be consider as an early attempt into what is know today as Multi-Task Learning (MTL)

   e. Tensorflow_models/feature_extraction_classification_models: different tensorflow implementations for SqueezeNet, Resnet, ShuffleNet, and MobileNet.

   f. caffe_models/ShuffleNet: Experimentation with shufflenet topology

   g. Tensorflow2Caffe_converter: Model converter from Tensorflow to Caffe.

2. computer_vision_img_vid: Different computer vision algorithms implemented on CPU and GPU for image and video in raw and compressed domain (H.264 standard), the folder is structured as follows:

   a. canny_edge_detection: Canny edge detection, fully implemented on GPU

   b. colormap_extractor: Color mapping extraction from RGB images

   c. data_augmentation: Python implemented data augmentation for input images

   d. gstreamer: further divided as follows:

    i. **gst_imgproc**: blob and skin detector on gstreamer, also moment normalization and color-retinex implementations on gstreamer.
   
    ii. **gst_rgb2gray**: RGB2GRAY implementation on gstreamer
   
    iii. **gst_rgbmapping**: Color mapping implementation for gstreamer
   

   e. hough_transform_lines_circles: Line and circle extraction using hough transform, fully implemented on GPU

   f. LBP_extract_module: LBP feature extraction from sample images

   g. spatio_temporal_saliency_maps: Static and Dynamic saliency mapping extraction from video/images

   h. ToneMapping: Color enhancement using Tone mapping algorithm

3. metaheuristic_algorithms: Implementation of several meta-heuristic algorithms including one developed during my master and PhD. degree. The folder is structured as follows:

   a. VOA: Virus Optimization Algorithm proposed for the first time in 2009 but accepted until 2014.

   b. TBD: more to be added in the future ...

4. miscellaneous: Subfolder containing different ideas tested over the past years, and do not have any specific field of application. This subfolder has the following structure:

   a. Bbox_filter: Bounding box filter for object detection algorithms (python)

   b. bitstream_analizer_openh264based: h264-bitstream saliency map extractor (C/C++)

   c. ffmpeg_video_handler_c: FFMPEG library based video/camera frame extractor (C/C++)

   d. fft_conformance: FFT conformance test to determine performance and accuracy (C/C++)

   e. gif_generator: GIF generator application (C/C++)

   f. h264_decoder_module_python: FFMPEG video decoder (python)

   g. hd5_rawImage_database_creator: HDF5 database image file generator (python)

   h. test_gstreamer_thread_priorities: Gstreamer thread priority test (C/C++)

   i. wavpack_gstreamer: Gstreamer Wavpack plugin encoder and file writer with Metadata Tags (C/C++)

5. caffe_own: Caffe repository with some modifications to support models and layers proposed over the past 5 years, for example: MobileNet, ShuffleNet, SSD, MaskRCNN, GAN, etc.

Contributors

Josue R. Cuevas

josuercuevas@gmail.com