docs/docs/configuration/object_detectors.md

@@ -365,8 +365,8 @@ detectors:
 
 model:
   model_type: rfdetr
-  width: 320
+  width: 560
-  height: 320
+  height: 560
   input_tensor: nchw
   input_dtype: float
   path: /config/model_cache/rfdetr.onnx
@@ -616,8 +616,8 @@ detectors:
 
 model:
   model_type: rfdetr
-  width: 320
+  width: 560
-  height: 320
+  height: 560
   input_tensor: nchw
   input_dtype: float
   path: /config/model_cache/rfdetr.onnx
@@ -983,21 +983,22 @@ Make sure you change the batch size to 1 before exporting.
 
 ### Download RF-DETR Model
 
-RF-DETR can be exported as ONNX by running the command below. You can copy and paste the whole thing to your terminal and execute, altering `MODEL_SIZE=Nano` in the first line to `Nano`, `Small`, or `Medium` size.
+To export as ONNX:
 
-```sh
+1. `pip3 install rfdetr`
-docker build . --build-arg MODEL_SIZE=Nano --output . -f- <<'EOF'
+2. `python3`
-FROM python:3.11 AS build
+3. `from rfdetr import RFDETRBase`
-RUN apt-get update && apt-get install --no-install-recommends -y libgl1 && rm -rf /var/lib/apt/lists/*
+4. `x = RFDETRBase()`
-COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
+5. `x.export()`
-WORKDIR /rfdetr
+
-RUN uv pip install --system rfdetr onnx onnxruntime onnxsim onnx-graphsurgeon
+#### Additional Configuration
-ARG MODEL_SIZE
+
-RUN python3 -c "from rfdetr import RFDETR${MODEL_SIZE}; x = RFDETR${MODEL_SIZE}(resolution=320); x.export()"
+The input tensor resolution can be customized:
-FROM scratch
+
-ARG MODEL_SIZE
+```python
-COPY --from=build /rfdetr/output/inference_model.onnx /rfdetr-${MODEL_SIZE}.onnx
+from rfdetr import RFDETRBase
-EOF
+x = RFDETRBase(resolution=560)  # resolution must be a multiple of 56
+x.export()
 ```
 
 ### Downloading YOLO-NAS Model

docs/docs/frigate/hardware.md

@@ -166,12 +166,16 @@ There are improved capabilities in newer GPU architectures that TensorRT can ben
 Inference speeds will vary greatly depending on the GPU and the model used.
 `tiny` variants are faster than the equivalent non-tiny model, some known examples are below:
 
-| Name            | YOLOv9 Inference Time | YOLO-NAS Inference Time   | RF-DETR Inference Time |
+| Name            | YOLOv7 Inference Time | YOLO-NAS Inference Time   | RF-DETR Inference Time    |
-| --------------- | --------------------- | ------------------------- | ---------------------- |
+| --------------- | --------------------- | ------------------------- | ------------------------- |
-| RTX 3050        | t-320: 15 ms          | 320: ~ 10 ms 640: ~ 16 ms | Nano-320: ~ 12 ms      |
+| GTX 1060 6GB    | ~ 7 ms                |                           |                           |
-| RTX 3070        | t-320: 11 ms          | 320: ~ 8 ms 640: ~ 14 ms  | Nano-320: ~ 9 ms       |
+| GTX 1070        | ~ 6 ms                |                           |                           |
-| RTX A4000       |                       | 320: ~ 15 ms              |                        |
+| GTX 1660 SUPER  | ~ 4 ms                |                           |                           |
-| Tesla P40       |                       | 320: ~ 105 ms             |                        |
+| RTX 3050        | 5 - 7 ms              | 320: ~ 10 ms 640: ~ 16 ms | 336: ~ 16 ms 560: ~ 40 ms |
+| RTX 3070 Mobile | ~ 5 ms                |                           |                           |
+| RTX 3070        | 4 - 6 ms              | 320: ~ 6 ms 640: ~ 12 ms  | 336: ~ 14 ms 560: ~ 36 ms |
+| Quadro P400 2GB | 20 - 25 ms            |                           |                           |
+| Quadro P2000    | ~ 12 ms               |                           |                           |
 
 ### ROCm - AMD GPU