Merge pull request #453 from roflcoopter/dev

YOLOv7 default
roflcoopter · Dec 31, 2022 · ed55649 · ed55649
2 parents 9a8d095 + 8624683
commit ed55649
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 21 deletions.
diff --git a/docker/amd64/Dockerfile.base b/docker/amd64/Dockerfile.base
@@ -47,9 +47,8 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
   sha256sum -c ww15.sum && \
   dpkg -i *.deb && \
   rm -R /opencl && \
-  rm /detectors/models/darknet/yolov7* && \
-  ln -s /detectors/models/darknet/yolov3.weights /detectors/models/darknet/default.weights && \
-  ln -s /detectors/models/darknet/yolov3.cfg /detectors/models/darknet/default.cfg
+  ln -s /detectors/models/darknet/yolov7.weights /detectors/models/darknet/default.weights && \
+  ln -s /detectors/models/darknet/yolov7.cfg /detectors/models/darknet/default.cfg
 
 # OpenVINO
 COPY --from=opencv /usr/local/runtime /usr/local/runtime/
diff --git a/docker/jetson-nano/Dockerfile.base b/docker/jetson-nano/Dockerfile.base
@@ -63,8 +63,8 @@ RUN \
   libegl1 \
   libfreetype6 \
   ffmpeg=${JETSON_NANO_FFMPEG_APT_VERSION} && \
-  ln -s /detectors/models/darknet/yolov4-tiny.weights /detectors/models/darknet/default.weights && \
-  ln -s /detectors/models/darknet/yolov4-tiny.cfg /detectors/models/darknet/default.cfg && \
+  ln -s /detectors/models/darknet/yolov7-tiny.weights /detectors/models/darknet/default.weights && \
+  ln -s /detectors/models/darknet/yolov7-tiny.cfg /detectors/models/darknet/default.cfg && \
   # Minimal cuda install does not create symlink so we do it manually
   ln -s /usr/local/cuda-10.2 /usr/local/cuda && \
   echo "/usr/lib/aarch64-linux-gnu/tegra" >> /etc/ld.so.conf.d/nvidia-tegra.conf && \

diff --git a/docs/src/pages/components-explorer/components/darknet/index.mdx b/docs/src/pages/components-explorer/components/darknet/index.mdx
@@ -15,20 +15,20 @@ import config from "./config.json";
 Darknet is a state-of-the-art object detector that uses the YOLO (You Only Look Once) framework.
 It is built on a singel-stage algorithm to achieve both speed and accuracy.
 
-YOLOv7 is currently the most accurate and fastest model.
+YOLOv7 is currently the most accurate and fastest model and has hardware acceleration support on both GPUs and CPUs.
 
 If CUDA is available on your system, `darknet` will run on your GPU.
 
 :::note
 
 `darknet` component uses the official [Darknet](https://github.com/AlexeyAB/darknet) implementation when running on a GPU.
-When running on a CPU, it uses OpenCV's implementation of Darknet, which has some limitations in what models can be used.
+When running on a CPU, it uses OpenCV's implementation of Darknet.
 
 :::
 
-:::warning YOLOv7
+:::info
 
-YOLOv7 is only available if you run the `roflcoopter/amd64-cuda-viseron` image.
+YOLOv7 is the default model used by `darknet` in all images.
 
 :::
 
@@ -122,18 +122,6 @@ It will help you find the perfect trade-off between accuracy and latency.
 
 </details>
 
-The default model differs a bit per container:
-
-| Image                             | Model                 |
-| --------------------------------- | --------------------- |
-| `roflcoopter/viseron`             | `yolov3.weights`      |
-| `roflcoopter/amd64-viseron`       | `yolov3.weights`      |
-| `roflcoopter/amd64-cuda-viseron`  | `yolov7.weights`      |
-| `roflcoopter/jetson-nano-viseron` | `yolov4-tiny.weights` |
-
-YOLOv4/YOLOv7 does not currently work in OpenCV which is why YOLOv3 is the default for some images.
-As soon as this is fixed for the versions of OpenCV that Viseron is using, YOLOv7 will be the standard for all.
-
 <Admonition type="tip">
 The containers also has <code>*-tiny.weights</code> model included in the image.
 The tiny-models can be used to reduce CPU and RAM usage.