AlphaDeep Capture

AlphaDeep Capture is a high-performance Python/Pygame application designed for live streaming, data capture, and realtime cloud-inference using the AlphaDeep AI platform.

Installation

Install the application directly from PyPI using pip:

pip install alphadeep-capture

Configuration

AlphaDeep Capture uses a YAML configuration file stored in your user configuration directory (following the XDG Base Directory standard).

Configuration Path: ~/.config/alphadeep/capture/config.yaml

If this file does not exist when the app is first launched, a default configuration will be automatically generated.

Example config.yaml

cameras:
  - id: 0
    name: Camera 1
  - id: 1
    name: Camera 2

fps: 30
samples_dir: ~/alphadeep_samples

# AlphaDeep API Configuration
alphadeep_session: "your_session_id"
alphadeep_adapter: "zero_adapter"
alphadeep_api_key: "your_api_key_here"

# Set the AI task. Supported values: 
# "detection", "segmentation", "classification", "ocr", "captioning"
alphadeep_task: "detection"

# (Optional) List of target classes to look for
alphadeep_classes: 
  - "car"
  - "person"

# Number of parallel threads to use when calling the AlphaDeep API (default is 4)
alphadeep_workers: 4

# Target square resolution to send to the server. Supported: 224, 448, 896
alphadeep_resolution: 448

# The default resolution for the camera hardware (if save_full_res is False) and standard saved images.
capture_resolution: [640, 480]

# Auto Capture configuration
autocapture:
  length: 2.0    # Total duration (in seconds) to run auto capture
  period: 0.5    # Trigger frequency (in seconds) between captures

# Enable verbose console output for debugging API responses
debug: true

Adding Your API Key

In order to use the Analyze functionality, you must edit your config.yaml and provide your alphadeep_api_key, alphadeep_session, and alphadeep_task.

Running the Application

Once installed, you can launch the application directly from your terminal:

alphadeep-capture

Features and Controls

• New Sample: Creates a fresh timestamped directory in your samples_dir for saving data.
• Capture: Instantly grabs the current frame from all active cameras and saves them as .jpg files in the current sample directory.
• Auto Capture: Automatically captures frames repeatedly over a configurable duration and frequency.
• Preview Sample: Opens an interactive gallery overlay to view all images captured in the current sample.
• Analyze: Sends the current sample to the AlphaDeep API for remote inference. Results are saved locally as .json and rendered interactively in the gallery (supports detection, segmentation, classification, OCR, and captioning).
• Comment: Opens a dialog to attach text notes to the current sample (saved as comment.txt).