Hailo Model Zoo
The Hailo Model Zoo provides pre-trained deep learning models plus a full environment for building, quantizing and evaluating them for deployment on Hailo AI accelerator devices.
The Hailo Model Zoo provides pre-trained models in ONNX and TensorFlow formats, along with pre-compiled HEF (Hailo Executable Format) binaries, for running on Hailo's AI accelerator hardware such as the Hailo-15H, Hailo-10H and Hailo-15L. It lets a user measure a model's full-precision accuracy, its quantized accuracy using the Hailo Emulator, and its accuracy when actually run on a Hailo device.
Models are grouped into public models, trained on publicly available datasets and covering tasks such as classification, object detection and segmentation, and models trained internally for specific use cases. The Zoo also provides re-training instructions so a model can be retrained on a custom dataset, and lets a user generate the HEF binary needed to accelerate a model on Hailo hardware.
It depends on specific pinned versions of Python, TensorFlow, CUDA and the Hailo Dataflow Compiler, and is released under the MIT license.