Abraia Vision SDK
The Abraia Vision SDK is a high-performance, edge-ready Python library and toolkit for computer vision, image processing, model training, and advanced inference. It unifies state-of-the-art vision models (such as YOLO, SAM, CLIP, and custom recognition pipelines) into a seamless API for production-ready applications, real-time video analysis, object tracking, hyperspectral imaging, and edge hardware deployment.
📚 Table of Contents
📦 Installation
Install the Abraia SDK from PyPI:
pip install -U abraia
For training and development run the installation with optional extras (dev, multiple):
pip install -U abraia[dev,multiple]
🚀 Core Modules & Features
1. Inference & Computer Vision (abraia.inference)
- Object Detection: Fast ONNX/YOLO-based object detection (
abraia.inference.Model). - Segmentation (SAM): Segment Anything Model integration for precise image masking (
abraia.inference.Sam). - Object Tracking & People Flow: Advanced multi-object tracking (
Tracker), line crossing counters (LineCounter), and region duration timers (RegionTimer). - Face Recognition: Identify and match faces in images and streams (
FaceRecognizer). - License Plate Recognition (ALPR): Automatic license plate detection and text recognition (
PlateRecognizer). - OCR: Extract text from images (
Ocr). - Semantic Search (CLIP): Vector embeddings and similarity search for text-to-image and image-to-image retrieval (
Clip).
2. Image Editing & Enhancement (abraia.editing)
- Upscaling: Super-resolution image enhancement (
upscale). - Smart Cropping: Intelligent content-aware cropping (
smartcrop). - Background Removal: Foreground segmentation and background removal (
removebg). - Inpainting: Image restoration and object removal (
inpaint).
3. Multispectral & Hyperspectral Imaging (abraia.multiple)
- Specialized tools for hyperspectral and multispectral image analysis, cube processing, and spectral signature extraction (
abraia.multiple.hsi).
4. Edge AI & Hardware Acceleration (abraia.hailo)
- Optimized runtime support and toolboxes for Hailo NPU hardware acceleration (
abraia.hailo).
5. Training & Dataset Operations (abraia.training)
- Tools for training custom classification (
classify) and detection (detect) models, along with dataset preprocessing utilities (dataset,ops).
6. Utilities & Video Processing (abraia.utils)
- Robust video frame iteration and manipulation (
Video). - Annotation and rendering tools (
render_results,render_counter,render_region). - Compression and sketch generation utilities.
💡 Examples & Usage Guides
People Monitoring & Tracking
Monitor people flow, count crossings, and track dwell times in public spaces or commercial areas:
from abraia.inference import Model, Tracker
from abraia.inference.tools import LineCounter, RegionTimer
from abraia.utils import Video, render_results, render_counter, render_region
model = Model("multiple/models/yolov8n.onnx")
video = Video('people-walking.mp4')
tracker = Tracker(frame_rate=video.frame_rate)
line_counter = LineCounter([(0, 650), (1920, 650)])
region_timer = RegionTimer([(10, 600), (1690, 600), (1690, 700), (10, 700)])
for k, frame in enumerate(video):
results = model.run(frame, labels=['person'])
results = tracker.update(results)
in_count, out_count = line_counter.update(results)
in_objects, out_objects = region_timer.update(results, k / video.frame_rate)
frame = render_counter(frame, line_counter.line, f"In: {in_count} | Out: {out_count}")
frame = render_region(frame, region_timer.region, f"Count: {len(in_objects)}")
frame = render_results(frame, in_objects)
video.show(frame)
Face Recognition
Identify and recognize people in images:
import os
from abraia.inference import FaceRecognizer
from abraia.utils import load_image, save_image, render_results
img = load_image('images/rolling-stones.jpg')
out = img.copy()
recognition = FaceRecognizer()
index = []
for src in ['mick-jagger.jpg', 'keith-richards.jpg', 'ronnie-wood.jpg', 'charlie-watts.jpg']:
img = load_image(f"images/{src}")
rslt = recognition.identify_faces(img)[0]
index.append({'name': os.path.splitext(src)[0], 'vector': rslt['vector']})
results = recognition.identify_faces(results, index)
render_results(out, results)
save_image(out, 'images/rolling-stones-identified.jpg')
License Plate Recognition (ALPR)
Automatically detect and recognize car license plates in images and video streams:
from abraia.inference import PlateRecognizer
from abraia.utils import load_image, show_image, render_results
alpr = PlateRecognizer()
img = load_image('images/car.jpg')
results = alpr.recognize(img)
frame = render_results(img, results)
show_image(img)
Semantic Search with CLIP
Search images using natural language text queries via CLIP embeddings:
from tqdm import tqdm
from glob import glob
from abraia.utils import load_image
from abraia.inference.clip import Clip
from abraia.inference.ops import search_vector
clip_model = Clip()
image_paths = glob('images/*.jpg')
image_index = [{'vector': clip_model.get_image_embeddings([load_image(image_path)])[0]} for image_path in tqdm(image_paths)]
text_query = "full body person"
vector = clip_model.get_text_embeddings([text_query])[0]
idxs, scores = search_vector(vector, image_index)
print(f"Similarity score is {scores[0]} for image {image_paths[idxs[0]]}")
🍓 Real-Time Edge Object Counter on Raspberry Pi with Hailo NPU
Deploy high-performance real-time object detection and counting on a Raspberry Pi equipped with a Hailo AI expansion board (such as Hailo-8 or Hailo-8L). This pipeline combines hardware-accelerated model inference (abraia.hailo), multi-object tracking (abraia.inference.Tracker), line crossing counters (LineCounter), and region timers (RegionTimer), integrated with the asynchronous video processing pipeline (VideoInput & VideoDisplay).
Implementation Guide
Create a script (e.g., edge_counter.py) ready for deployment on your Raspberry Pi:
import threading
from abraia.hailo.toolbox import ModelInference
from abraia.inference import Tracker
from abraia.inference.tools import LineCounter, RegionTimer
from abraia.utils import VideoInput, VideoDisplay, render_results, render_counter, render_region
from abraia.hailo.detect import run_inference_pipeline
# 1. Initialize threaded video input (e.g., Raspberry Pi Camera or RTSP stream)
stop_event = threading.Event()
input_data = VideoInput(input_src=0, resolution=(1920, 1080), stop_event=stop_event)
visualizer = VideoDisplay(source_fps=input_data.source_fps, stop_event=stop_event)
# 2. Load Hailo compiled model (.hef) optimized for edge NPU
model_inference = ModelInference(
hef_path="yolov8n.hef",
task="detect",
labels=["person", "car"],
batch_size=1,
score_threshold=0.3
)
# 3. Setup Tracker & Analytics Tools (Line Counter & Region Timer)
tracker = Tracker(frame_rate=input_data.source_fps or 30.0)
line_counter = LineCounter([(100, 540), (1820, 540)]) # Crossing boundary line
region_timer = RegionTimer([(300, 200), (1620, 200), (1620, 900), (300, 900)]) # Zone of interest
# 4. Custom Inference & Analytics Result Handler
def edge_processing_handler(frame, detections, tracker=None, tracklet_history=None):
if tracker:
detections = tracker.update(detections)
# Update line crossing and region analytics
in_count, out_count = line_counter.update(detections)
in_objects, out_objects = region_timer.update(detections, 1.0 / (input_data.source_fps or 30.0))
# Render real-time visual overlays
frame = render_counter(frame, line_counter.line, f"In: {in_count} | Out: {out_count}")
frame = render_region(frame, region_timer.region, f"Zone Count: {len(in_objects)}")
return render_results(frame, detections)
# 5. Run High-Performance Edge Pipeline
try:
run_inference_pipeline(
model_inference=model_inference,
input_data=input_data,
visualizer=visualizer,
tracker=tracker
)
finally:
stop_event.set()
Deployment on Raspberry Pi
Execute the script directly on the Raspberry Pi:
python3 edge_counter.py
📄 License
This project is licensed under the MIT License.
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
abraia-0.27.3.tar.gz
(88.0 kB
view details)
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file abraia-0.27.3.tar.gz.
File metadata
- Download URL: abraia-0.27.3.tar.gz
- Upload date:
- Size: 88.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
9ef5b2c97b143efd014ef4d7a4422a3e5e20a0bb97cf4ccfda09066dbfaecfc8
|
|
| MD5 |
c4a16dcaf8d95dce6a663ce8cbccb218
|
|
| BLAKE2b-256 |
bcf6021df363b9c41e364e824d7afe30463aa12d0399ab290c1e8d08d36117b0
|
File details
Details for the file abraia-0.27.3-py3-none-any.whl.
File metadata
- Download URL: abraia-0.27.3-py3-none-any.whl
- Upload date:
- Size: 1.5 MB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f0372bb5cd2fd305107dd826efe408c37aed0f05fc6da86ac13663e41f3532c0
|
|
| MD5 |
9875a2026d60a94443d340987e9091c2
|
|
| BLAKE2b-256 |
8e97d243179057796b6de838328362908f3c7acc56a869d76a0b2f6739fa5240
|
