Update RT-DETR and YOLOX

README.md

@@ -222,6 +222,7 @@ NVIDIA DeepStream SDK 7.1 / 7.0 / 6.4 / 6.3 / 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 /
 * [DAMO-YOLO](https://github.com/tinyvision/DAMO-YOLO)
 * [PP-YOLOE / PP-YOLOE+](https://github.com/PaddlePaddle/PaddleDetection/tree/develop/configs/ppyoloe)
 * [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md)
+* [RT-DETR](https://github.com/lyuwenyu/RT-DETR)
 
 ##
 

config_infer_primary_yolox.txt

@@ -1,7 +1,7 @@
 [property]
 gpu-id=0
 net-scale-factor=1
-model-color-format=0
+model-color-format=1
 onnx-file=yolox_s.onnx
 model-engine-file=model_b1_gpu0_fp32.engine
 #int8-calib-file=calib.table

docs/RTDETR_Paddle.md

@@ -141,6 +141,15 @@ maintain-aspect-ratio=0
 ...
 ```
 
+**NOTE**: The **RT-DETR** do not require NMS. To get better accuracy, use
+
+```
+[property]
+...
+cluster-mode=4
+...
+```
+
 ##
 
 ### Edit the deepstream_app_config file

docs/RTDETR_PyTorch.md

@@ -160,6 +160,15 @@ maintain-aspect-ratio=0
 ...
 ```
 
+**NOTE**: The **RT-DETR** do not require NMS. To get better accuracy, use
+
+```
+[property]
+...
+cluster-mode=4
+...
+```
+
 ##
 
 ### Edit the deepstream_app_config file

docs/RTDETR_Ultralytics.md

@@ -160,6 +160,15 @@ maintain-aspect-ratio=0
 ...
 ```
 
+**NOTE**: The **RT-DETR Ultralytics** do not require NMS. To get better accuracy, use
+
+```
+[property]
+...
+cluster-mode=4
+...
+```
+
 ##
 
 ### Edit the deepstream_app_config file

docs/YOLOX.md

@@ -143,6 +143,24 @@ symmetric-padding=0
 ...
 ```
 
+**NOTE**: The **YOLOX** uses BGR color format for the image input. It is important to change the `model-color-format` according to the trained values.
+
+```
+[property]
+...
+model-color-format=1
+...
+```
+
+**NOTE**: The **YOLOX legacy** uses RGB color format for the image input. It is important to change the `model-color-format` according to the trained values.
+
+```
+[property]
+...
+model-color-format=0
+...
+```
+
 **NOTE**: The **YOLOX** uses no normalization on the image preprocess. It is important to change the `net-scale-factor` according to the trained values.
 
 ```

nvdsinfer_custom_impl_Yolo/yolo.cpp

@@ -146,9 +146,9 @@ Yolo::createEngine(nvinfer1::IBuilder* builder)
             "config_infer file to get better accuracy\n" << std::endl;
     }
   }
-  if (m_ClusterMode != 2) {
-      std::cout << "NOTE: Wrong cluster-mode is set, make sure to set cluster-mode=2 on the config_infer file\n" <<
-          std::endl;
+  if (m_ClusterMode != 2 && m_ClusterMode != 4) {
+      std::cout << "NOTE: Wrong cluster-mode is set, make sure to set cluster-mode=4 (RT-DETR or custom NMS) or " <<
+          "cluster-mode=2 on the config_infer file\n" << std::endl;
   }
 
   if (m_NetworkMode == "FP16") {