cucafly Goto Github PK

augmented-random-search-with-python-openai-and-gym

Here is an implementation of a new type of artificial intelligence, which is almost as powerful as the algorithm used by Google Deep Mind to train an AI to walk and run through an environment! The name of this technique is Augmented Random Search, it was created in 2018 and is on average 15 times faster than traditional algorithms! This algorithm is within the Reinforcement Learning area, which is a type of learning used in multi-agent systems in which agents must interact in the environment and learn on their own, earning positive rewards when they perform correct actions and negative rewards when they perform actions. that don't lead to the goal. Artificial intelligence learns without any prior knowledge, adapting to the environment and finding solutions alone! The application consists of training a simulation of a robot that needs to learn to walk in an environment. We will use the ARS (Augmented Random Search) technique, Python as a programming language and the OpenAI Gym as a simulation environment.

automatic-character-generation-with-gan-mnist

The content refers to an example application of GAN (Generative Adversarial Networks). The goal is to generate characters automatically. The database used for training the neural network was MNIST (handwritten digit data). The practical example was implemented in Jupyter Notebook. I used TensorFlow 2.0 and Python 3.

badgr

counter-of-fingers-with-python-in-jupyter-lab

Python Algorithm running on Jupyter Lab for finger recognition and counting.

deep-learning-on-point-cloud-for-3d-classification-and-segmentation

deteccao-de-emocoes-em-videos

Deep Leaning algorithm with Convolutional Neural Network to recognize some emotions in photos and videos.

detection-of-pistol-by-deep-learning-with-yolo_v5

Construction of a lumbar weapon detection system through CCTV cameras with the help of artificial intelligence neural network (YOLO_v5), which has been trained by the private dataset of Pashtaw.

drl-robot-navigation

Deep Reinforcement Learning for mobile robot navigation in ROS Gazebo simulator. Using Twin Delayed Deep Deterministic Policy Gradient (TD3) neural network, a robot learns to navigate to a random goal point in a simulated environment while avoiding obstacles.

forecast-of-the-stock-exchange-price-with-recurring-neural-networks

The content consists of an example of application of recurrent neural networks (Recurring Neural Network- RNN) using using a database in csv, sklearn library and python 3. The objective is to predict the price of the stock exchange, a typical case of time series suitable for RNN application.

gymnasium

A standard API for reinforcement learning and a diverse set of reference environments (formerly Gym)

oidv6_toolkit_download_open_images_support_yolo_format

Download and visualize single or multiple classes from the huge Open Images v6 dataset

reconhecimento-de-libras-usando-python-e-opencv

Hand - Translating Libras into images For the more dynamic identification of various positions, 4 modules were developed to extract some characteristics and compare the displacement of the joints (key points). We extract HEIGHT, POSITION and PROXIMITY between the detected key points. Module extrator_ALTURA: checks if a point is 'above' or 'below' another specific point. For example, if the fingertip is 'above' the wrist. Extrator_POSICAO module: functions to check if the fingers are 'bent' or 'stretched', horizontally or vertically. It also receives the result of the extractor_ALTURA module to find out what position the hand is in (turned 'up' or 'down'). Extractor_PROXIMITY module: functions that compare the proximity between the detected key points. For example, if the result of the extractor_POSICAO module is equal to 'folded' for the index and middle fingers and both are at the same height, then it means that the fingers are close together. Alphabet module: after extracting all these characteristics, the characteristics alphabet was created, where a VECTOR OF VECTORS receives the result of all the extractor modules. So we use this module to compare with a new analysis (new image) of input. The letters: H, J, K, X,, Y, Z were not used due to the additional movement for the correct execution of the letters. These letters can be analyzed in a different function, similar to the body position analysis function, where we compare the transition between points. Letter T: for the thumb, due to being overlapped by the index finger, the algorithm does not recognize the points on the fingertip. Letter N and U are confused.

reconhecimento-de-polichinelo-em-videos

Deep Learning algorithm with Convolutional Neural Network to detect and recognize the correct movement of an exercise type.

reduced-image-dimension-with-autoencoders

The content is a simple example of the great importance of using autoencoders in complex projects that involve large image data. The example was implemented in Jupyter Notebook in Python 3.

ros2djs

2D Visualization Library for use with the ROS JavaScript Libraries

virtual-autonomous-car-with-artificial-intelligence-and-deep-q-learning

Construction of a virtual autonomous car using reinforcement learning. We will use modern Deep Learning techniques with the PyTorch library and the Python language.

web-based-navigation-with-rosbridge

Motivation: Cloud robotics is an emerging topic in the last few years and it aims at integrating mobile robots with cloud computing systems and the Internet of Things through web services interfaces. Project Overview: use of React and JavaScript to control and monitor robot ROS navigation using Web interfaces. This didactic project comes from the series of courses "ROS for begineers" by professor and doctor Anis Koubaa. What was developed in this project: create a web interface to teleoperate a robot, develop a web interface for robot navigation, use JavaScript React front-end development and use ROSBridge to interact with ROS ecosystem.

yolov3

Keras implementation of yolo v3 object detection.

yolov7

Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

yolov7_d2

🔥🔥🔥🔥 (Earlier YOLOv7 not official one) YOLO with Transformers and Instance Segmentation, with TensorRT acceleration! 🔥🔥🔥