Project structure of: MarkFzp/act-plus-plus

• act-plus-plus Advanced Robotics & AI Tools
  • __init__.py Incomplete or empty code snippet
  • align.py Calibrates head cam and arms, wakes bots, opens grippers
  • commands.txt Trains RL models, preprocesses data, experiments with hyperparameters.
  • compress_data.py Compresses images, videos, and handles HDF5 datasets.
  • conda_env.yaml Conda environment setup: "aloha" with packages.
  • constants.py Robotics task parameters and environments defined.
  • detr DETR data utilities and visualization.
    • main.py Deep learning transformer detector script.
    • models DETR: Transformer models for images, sequences.
      • __init__.py DETR-VAE and CNN+MLP model build functions in one module.
      • backbone.py Backbone Class: ResNet, Vision Transformer
      • detr_vae.py DETR-VAE: Transformer-based image detection model with CVAE
      • latent_model.py Latent Model Transformer for sequence modeling
      • position_encoding.py Sine-cosine positional embedding transformer class.
      • transformer.py PyTorch Transformer class with encoder, decoder
    • README.md DETR: End-to-End Object Detection with Transformers
    • setup.py Setup script for detr package with setuptools.
    • util Data processing and visualization utilities for DETR.
      • __init__.py Facebook owns the codebase rights.
      • box_ops.py Box operations and GIoU functions.
      • misc.py Track, smooth, and log metrics with progress updates.
      • plot_utils.py Generate matplotlib plots using training logs and precision-recall curves.
  • dxl_test.py Python script connects motors, prints info
  • dynamixel_client.py Dynamixel motor control with DynamixelSDK and CLI.
  • ee_sim_env.py Bi-manual robot environment sim, tasks and rewards.
  • imitate_episodes.py Trains robot policy network with RL, VQ-VAE, logs data, saves checkpoints, validates performance.
  • policy.py Policy network for multi-camera image tasks, RL with normalization and loss.
  • postprocess_episodes.py Post-process episodes: loads, processes, scales, compresses, and saves data.
  • README.md Comprehensive RL implementation and guides.
  • record_sim_episodes.py Simulates environment, stores data for analysis.
  • replay_episodes.py Replay episodes: Import, organize, save videos.
  • scripted_policy.py BasePolicy class for robotic arm policy with trajectory generation and gripper commands.
  • setup.py Create Python distribution package with setup script.
  • sim_env.py Interactive 3D simulator for cube manipulation tasks.
  • train_actuator_network.py Trains actuator network, visualizes predictions.
  • train_latent_model.py Train latent model for robot manipulation using ACT-Plus-Plus and deep reinforcement learning.
  • truncate_data.py Truncates dataset, creates observation group with limited image data.
  • utils.py EpisodicDataset: Data Processing, Augmentations, Torch Tensor Compat
  • vinn_cache_feature.py Cache VINN feature extraction.
  • vinn_eval.py Nearest neighbor calculation and rollouts for image classification.
  • vinn_select_k.py Optimize k for nearest neighbor losses from datasets.
  • visualize_episodes.py Visualize timestamped camera frames in Python.