EdgeFirst Validator

EdgeFirst Validator is a Python CLI tool for validating computer vision models (object detection, instance segmentation, semantic segmentation, multitask) against annotated datasets. It supports multiple inference backends — ONNX, TFLite, Keras, TensorRT, DeepViewRT, and Kinara Ara2 — and integrates with EdgeFirst Studio for cloud and on-target validation workflows. Published to PyPI as edgefirst-validator.

Overview

• Changelog
• Installation
• Usage
• Architecture
• Testing Results
• Documentation

Changelog

• Mar 10, 2026 [v5.5.0]: HAL decode_masks() migration, process_masks_retina, edgefirst-hal optional dependency, artifact download fix.
• Feb 17, 2026 [v5.4.0]: Split tensor decoding, cached dataset validation.
• Jan 28, 2026 [v5.3.0]: YOLO26 support and GREY color space support
• Jan 20, 2026 [v5.2.0]: Migrate color conversions to edgefirst-cameraadaptor library.
• Dec 19, 2025 [v5.1.0]: Handle metadata embedded in the model. Integrate changes to EdgeFirst HAL (edgefirst_hal)
• Nov 26, 2025 [v5.0.0]: Merge Ultralytics and Deployment (EdgeFirst) metrics. Computation of deployment metrics @ optimal thresholds. Defined validation stages.
• Nov 14, 2025 [v4.2.0]: Added HAL support (edgefirst_python). Added Kinara model support. Distinguish semantic and instance segmentation metrics.
• Aug 24, 2025 [v4.1.0]: Added ModelPack external decoding. Added Keras support. Added TP vs FP scores chart. Removed +1 padding to the NumPy NMS.
• June 16, 2025 [v4.0.0]: EdgeFirst Validator (formerly Deep View Validator) - Ultralytics metrics and EdgeFirst Studio integrations.

Installation

For on target validation, the system may already contain the dependencies needed to run the model such as tflite_runtime for TensorFlow Lite. In such cases, it is recommended to first create a python virtual environment with the --system-site-packages option enabled.

EdgeFirst-Validator is available in PyPI and can be installed via pip.

pip install edgefirst-validator

Validation of specific models requires these installations which includes the packages needed to run the models, in case the system at present does not contain these packages.

• HAL (GPU-accelerated preprocessing and mask decoding): pip install edgefirst-validator[hal]
• Keras: pip install edgefirst-validator[keras]
• ONNX: pip install edgefirst-validator[onnx]
• TensorRT: pip install edgefirst-validator[tensorrt]
• TFLite: pip install edgefirst-validator[tflite]
• Studio (all backends for cloud validation): pip install edgefirst-validator[studio]

Usage

1. As a quickstart, deploy default validation of a YOLOv5 model with COCO128 dataset.

   edgefirst-validator
   

2. Deploy validation as a user-managed session in EdgeFirst Studio. First login with EdgeFirst Client.

   edgefirst-client login
   

   edgefirst-validator --session-id <validator session ID>
   

   Otherwise, you can specify the credentials directly.

   edgefirst-validator --token <token> --session-id <validator session ID>
   

   edgefirst-validator --username <username> --password <password> --session-id <Validator Session ID>
   

3. Deploy standalone validation.

   edgefirst-validator <model path> <dataset path>
   

Documentation

Run the following commands to generate the documentation.

cd doc
pip install -r requirements.txt

make clean
make html

Unit Tests

Run the following commands to run the unit tests

pip install -r test/requirements.txt

python -m pytest

Build Instructions

Build the python wheel with these commands

pip install build
python -m build

Build the AMI using Packer

1. Install packer on Ubuntu

   sudo apt-get update
   sudo apt-get install -y gnupg software-properties-common curl
   

   Add HashiCorp repo:

   curl -fsSL https://apt.releases.hashicorp.com/gpg | sudo gpg --dearmor -o /usr/share/keyrings/hashicorp.gpg
   echo "deb [signed-by=/usr/share/keyrings/hashicorp.gpg] https://apt.releases.hashicorp.com $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/hashicorp.list
   

   sudo apt-get update
   sudo apt-get install -y packer
   

2. Initialize packer

   packer init validator.pkr.hcl
   

3. Build AMI

   packer build -var "aws_access_key=$AWS_ACCESS_KEY_ID" -var "aws_secret_key=$AWS_SECRET_ACCESS_KEY" validator.pkr.hcl