Update rendaring

README.md

@@ -71,7 +71,7 @@ $ make -j
 | Model | Resolutions | GFLOPS | Params | Precision | Sparsity | DNN time on RTX3080 | DNN time on Jetson Orin NX 16GB GPU | DNN time on Jetson Orin NX 16GB DLA| DNN time on Jetson Orin Nano 4GB GPU | cfg | weights |
 |---|---|---|---|---|---|---|---|---|---|---|---|
 | lightNet | 1280x960 | 58.01 | 9.0M | int8 | 49.8% | 1.3ms | 7.6ms | 14.2ms | 14.9ms | [github](https://github.com/daniel89710/lightNet/blob/master/cfg/lightNet-BDD100K-1280x960.cfg) |[GoogleDrive](https://drive.google.com/file/d/1qTBQ0BkIYqcyu1BwC54_Z9T1_b702HKf/view?usp=sharing) |
-| LightNet+Semseg | 1280x960 | 76.61 | 9.7M | int8 | 49.8% | 2.06ms | 15.3ms | 23.2ms | 26.2m | [github](https://github.com/daniel89710/lightNet-TRT/blob/main/configs/lightNet-BDD100K-det-semaseg-1280x960.cfg) | [GoogleDrive](https://drive.google.com/file/d/1ttdVtlDiPun13EQCB4Nyls3Q8w5aXg1i/view?usp=sharing)|
+| LightNet+Semseg | 1280x960 | 76.61 | 9.7M | int8 | 49.8% | 2.06ms | 15.3ms | 23.2ms | 26.2ms | [github](https://github.com/daniel89710/lightNet-TRT/blob/main/configs/lightNet-BDD100K-det-semaseg-1280x960.cfg) | [GoogleDrive](https://drive.google.com/file/d/1ttdVtlDiPun13EQCB4Nyls3Q8w5aXg1i/view?usp=sharing)|
 
  * "DNN time" refers to the time measured by IProfiler during the enqueueV2 operation, excluding pre-process and post-process times.
  * Orin NX has three independent AI processors, allowing lightNet to be parallelized across a GPU and two DLAs.																

modules/class_yolo_detector.hpp

@@ -63,7 +63,7 @@ class YoloDectector
     uint32_t *colormap = _p_net->get_detection_colormap();
     int id = result.id;
     if (colormap) {
-      cv::rectangle(img, result.rect, cv::Scalar(colormap[3*id+2], colormap[3*id+1], colormap[3*id+0]), 2);
+      cv::rectangle(img, result.rect, cv::Scalar(colormap[3*id+2], colormap[3*id+1], colormap[3*id+0]), 4);
     } else {
       cv::rectangle(img, result.rect, cv::Scalar(255, 0, 0), 2);
     }

modules/yolo.cpp

@@ -347,11 +347,15 @@ void Yolo::createYOLOEngine(const nvinfer1::DataType dataType, Int8EntropyCalibr
 	  std::string softmaxLayerName = "segmenter_" + std::to_string(segmenter_count);
 	  softmax->setName(softmaxLayerName.c_str());
 	  previous = softmax->getOutput(0);
+
+
+
 	  previous->setName(softmaxLayerName.c_str());
 	  assert(previous != nullptr);
 	  m_Network->markOutput(*previous);
 	  std::string outputVol = dimsToString(previous->getDimensions());
 	  tensorOutputs.push_back(softmax->getOutput(0));
+	  //tensorOutputs.push_back(previous);	  
 	  printLayerInfo(layerIndex, "segmenter", inputVol, outputVol, "    -");
 
 	  TensorInfo& curYoloTensor = m_OutputTensors.at(outputTensorCount);
@@ -792,9 +796,10 @@ std::vector<cv::Mat> Yolo::get_depthmap(std::vector<cv::Mat> &argmax) {
 	for (int x = 0; x < tensor.grid_w; x++) {
 	  int id = gray.at<unsigned char>(y, x);
 	  float rel = id/(float)c;
-	  int tmp = 120 + 90 * (1.0-rel);
+	  //int tmp = 120 + 90 * (1.0-rel);
 	  //tmp = tmp > 60.0 ? 60.0 : tmp;
-	  int hue = tmp > 180 ? tmp-180 : tmp;
+	  int hue = 120 + 90 * (1.0-rel);
+	  hue = hue > 180 ?  (180-hue) * (-1) : hue;
 	  unsigned char val = 255 - 180 * (1.0-rel);
 	  //hue = hue < 300 ? 0 : hue;
 	  hsv.at<cv::Vec3b>(y, x)[0] = hue;