trism-cv 0.0.1.dev0

A lightweight and modular Python package for handling computer vision inference (image/video) with Triton Inference Server.

TRISM Inference Script

This script performs batch inference on a folder of images using a Triton Inference Server model.

📂 Input

• image_folder: Path to the folder containing images (jpg, png, etc.).
• Images are loaded using OpenCV (cv2.imread) into a list of np.ndarray objects.

⚙️ Configuration

image_folder = "path/to/image_folder"   
model_name = "yolov_deyo_ensemble"
batch_size = 1 # You can pass a custom batch size or use default (e.g., 1)

🚀 Inference

model = TritonModel(
    model=model_name, 
    version=1,                    # Model version on Triton server
    url="localhost:8001",         # Triton server address
    grpc=True                     # Use gRPC protocol for communication
)

outputs = model.run(
    data_list=images,              # list of images 
    auto_config=True,
    batch_size=batch_size
)

📤 Output

• A list of numpy arrays, one for each input image.
• Each output has shape (n_detections, 6) where 6 = [x1, y1, x2, y2, confidence, class_id]

🧪 Debug Output

for i, out in enumerate(outputs):
    print(f"Image {i}: shape = {out.shape}, dtype = {out.dtype}")

License

GNU AGPL v3.0.
Copyright © 2025 Tien Nguyen Van. All rights reserved.