Embedl-optimized
- torchvision/densenet121
- INT8
- Orin Nano Super
- Labels
- Image
- Classification
- Build
- Embedl-optimized
- Precision
- INT8
- Runtime
- TensorRT with Embedl optimizations
DenseNet-121 is an image classifier from torchvision. It labels an image with one of the 1,000 ImageNet classes.
Methodology
- Latency: on the NVIDIA Jetson Orin Nano Super, batch 1, median GPU compute time (no copies to and from the GPU), with the engine compiled on the device from the build's quantized ONNX.
- Accuracy: the ImageNette validation split, 3,925 images with ImageNet-1k labels, measured on the NVIDIA Jetson Orin Nano Super.
- Size: the serialized TensorRT engine.
- Baseline TensorRT: the original model in FP16 on upstream TensorRT, same settings, same device. Not measured yet.
- Build: Embedl Deploy quantization (INT8 weights per channel and activations), calibrated on 5,000 ImageNet images, exported to ONNX and compiled with Embedl's TensorRT transforms.