Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow

KO | EN

gitlite — search

#javascript #react #ai #go #golang #machine-learning #rust #nodejs #macos #java #vue #pytorch #ios #open-source

Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow

(★ 11)

Github del curso de Udemy Deep Learning para Computer Vision con Python y TensorFlow

Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Deep Learning para Computer Vision con Python y TensorFlow

## Overview

This repository is a learning-oriented collection of notebooks, scripts, sample images, and presentation material for computer vision and deep learning with Python and TensorFlow. The content is organized as course-style examples rather than a packaged application or production service.

The examples cover image preprocessing, convolutional neural networks, transfer learning, object detection, OCR, image search, clustering, interpretability, reinforcement learning, and video classification workflows. Several notebooks were written for notebook platforms such as Kaggle or Google Colab and reference datasets that are not fully included in this repository.

## Key Topics

- Image preprocessing with OpenCV and NumPy.
- CNN training and evaluation with TensorFlow/Keras.
- Transfer learning using pretrained image feature extractors.
- TensorBoard logging for model training experiments.
- HOG feature extraction and dense neural network classification on MNIST.
- Object detection data preparation, TFRecord generation, model training, export, and inference using TensorFlow Object Detection API tooling.
- Image feature extraction with ResNet101 and nearest-neighbor image search.
- K-means clustering over deep image features.
- Simple Q-learning and a deep Q-network notebook for reinforcement learning examples.
- Video violence detection using frame-level features, TFRecords, and a 1D CNN.
- Basic model interpretability examples for image classification.

## Tech Stack

- Python
- Jupyter Notebook
- TensorFlow and Keras
- TensorFlow Hub
- TensorFlow Object Detection API tooling
- OpenCV
- NumPy
- Matplotlib
- scikit-learn
- NLTK clustering utilities
- Gym, keras-rl2, and Atari environments for reinforcement learning notebooks
- TensorBoard logs for selected experiments

There is no `requirements.txt`, `environment.yml`, `pyproject.toml`, or Docker configuration in the repository, so dependency versions are not pinned.

## Repository Structure

```text
deepLearningCV/
  Presentaciones/
    Machine Learning/
    Procesamiento de imagenes/
  codigo/
    Deep Clustering/
    Deep Reinforcement Learning/
    Deteccion Violencia en Video/
    DL Interpretable/
    ObjectDetection/
    Transfer Learning/
    facial expression/
    image search/
    lunares/
    ocr/
    procesamiento de imágenes/
    tumores/
```

Important files and folders:

- `deepLearningCV/codigo/tumores/detecciontumor-cnn.ipynb`: CNN training, prediction, evaluation, and TensorBoard workflow for tumor image classification.
- `deepLearningCV/codigo/Transfer Learning/transferlearningcnn.ipynb`: image classification with a custom CNN and transfer learning.
- `deepLearningCV/codigo/ocr/OCR_RedNeuronal.ipynb`: MNIST OCR example using HOG features and a neural network classifier.
- `deepLearningCV/codigo/ObjectDetection/notebook8ba14b6184.ipynb`: TensorFlow Object Detection API workflow for playing card detection.
- `deepLearningCV/codigo/ObjectDetection/empaquetadoobjectdetection.ipynb`: notebook steps for packaging TensorFlow object detection code.
- `deepLearningCV/codigo/image search/pokemonevol.ipynb`: image feature extraction and nearest-image search.
- `deepLearningCV/codigo/Deep Clustering/dclustering.ipynb`: feature extraction and K-means clustering over image embeddings.
- `deepLearningCV/codigo/Deteccion Violencia en Video/deteccionviolenciavideo.ipynb`: video classification pipeline using extracted frame features and TFRecords.
- `deepLearningCV/codigo/Deep Reinforcement Learning/Q_Learning.py`: compact Q-learning example on a small grid environment.
- `deepLearningCV/codigo/Deep Reinforcement Learning/deepqnet.ipynb`: DQN example using a Space Invaders environment.
- `deepLearningCV/codigo/lunares/clasificadorLunares.py`: OpenCV and NumPy-based color feature classifier for mole image examples.
- `deepLearningCV/codigo/tumores/logs/`: TensorBoard event files from a CNN training run.

## Technical Approach

Most examples follow a notebook-first workflow:

1. Load image, video, or environment data.
2. Preprocess inputs with OpenCV, NumPy, or TensorFlow data utilities.
3. Build features directly, extract deep features from pretrained CNNs, or define a TensorFlow/Keras model.
4. Train or run inference in the notebook.
5. Evaluate results with accuracy, loss, confusion matrices, TensorBoard logs, plots, or visualized predictions.

The repository does not define a shared Python package, reusable training CLI, central configuration system, or automated test suite. Each notebook or script is intended to be studied and run independently.

## Getting Started

Because dependencies are not pinned, create an isolated Python environment before running examples:

```bash
python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
```

Install the libraries required by the notebook you plan to run. A practical base environment for many examples is:

```bash
pip install jupyter numpy matplotlib opencv-python tensorflow scikit-learn nltk tensorflow-hub
```

Some notebooks require additional packages or platform-specific setup:

- Object detection notebooks expect TensorFlow Object Detection API tooling and Kaggle-style input paths.
- Reinforcement learning notebooks may require `gym`, `keras-rl2`, Atari ROM setup, and compatible TensorFlow/Keras versions.
- Some notebooks use Google Colab-specific helpers such as `google.colab.patches`.
- Several notebooks reference external Kaggle datasets. Download or mount those datasets and update paths before running locally.

Start Jupyter from the repository root:

```bash
jupyter notebook
```

Then open the target notebook under `deepLearningCV/codigo/`.

## Usage

Run each notebook independently and adapt dataset paths as needed. The notebooks include their own data loading, preprocessing, model definition, training, and evaluation cells.

For the standalone Q-learning example:

```bash
python "deepLearningCV/codigo/Deep Reinforcement Learning/Q_Learning.py"
```

For `deepLearningCV/codigo/lunares/clasificadorLunares.py`, update the dataset paths in the script before running it locally. The included `datasetLunares/` folder contains small class-specific image examples, but the script currently expects local absolute paths from the original authoring environment.

## Configuration

No environment variables or secret-based configuration files are required by the repository as committed.

Dataset paths are hardcoded in several notebooks and scripts. Treat those paths as placeholders for local, Kaggle, or Colab-mounted datasets and replace them with paths available in your own environment.

## Results and Outputs

The repository includes executed notebook outputs and TensorBoard event files for at least one tumor CNN experiment under `deepLearningCV/codigo/tumores/logs/`. Other expected outputs are generated within notebooks, including:

- Training and validation metrics.
- Model evaluation results.
- Confusion matrices.
- Image plots and visualized detections.
- TFRecord files and exported model artifacts when running the object detection and video notebooks.

Generated model checkpoints, exported models, and large datasets are not consistently included in the repository.

## Limitations

- This is a course/demo repository, not a production-ready ML package.
- Dependency versions are not pinned, so some notebooks may require updates for current TensorFlow, Keras, Gym, or Object Detection API versions.
- Several notebooks rely on Kaggle or Colab paths and external datasets that are not part of the repository.
- Some scripts contain hardcoded local paths that must be changed before local execution.
- There is no automated test suite, CI configuration, or reproducible environment file.
- Notebooks are organized by topic, but there is no single end-to-end application entry point.

## Future Improvements

- Add a `requirements.txt` or `environment.yml` with tested dependency versions.
- Replace hardcoded dataset paths with documented path variables or notebook parameters.
- Add lightweight setup notes per notebook, including required datasets.
- Move reusable preprocessing and feature extraction helpers into small Python modules.
- Add a short results table for completed experiments where metrics are available.
- Add `.gitignore` rules for generated checkpoints, TFRecords, and exported model artifacts.

Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow

Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow

Deep-Learning-para-Computer-Vision-con-Python-y-TensorFlow 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README