326 bug fix for gr spectrumdetect

gr-spectrumdetect/Dockerfile

@@ -9,6 +9,7 @@ RUN apt-get update && apt-get install -y \
     python3 \
     python3-dev \
     python3-numpy \
+    python3-apt \
     pip \
     gcc \
     g++ \
@@ -57,7 +58,7 @@ RUN pip install torchaudio -U
 
 RUN pip install pybombs
 
-RUN pybombs config makewidth 56
+RUN pybombs config makewidth 4
 
 RUN pybombs recipes add gr-recipes https://github.com/gnuradio/gr-recipes.git
 

gr-spectrumdetect/README.md

@@ -1,16 +1,14 @@
-## GR-Spectrumdetect Overview
+## gr-spectrumdetect Overview
 ---
 gr-spectrumdetect is an open-source example of using a trained model from TorchSig Wideband with GNU Radio for RF energy detection. 
 
 `detect.pt` can be downloaded with the bash `trained_model_download.sh` script in `gr-spectrumdetect/examples/`.
-- If `trained_model_download.sh` does not work, try using `trained_model_download_github.sh` (Downloads the model from [TorchSig's Github v0.6.0 Release Notes](https://github.com/TorchDSP/torchsig/releases/tag/v0.6.0))
 
 `detect.pt` is a YOLOv8x model trained for detection with the following settings:
 - single_cls=True
 - 1024x1024 spectrograms
 - gray scale black hot images
-- Wideband with level 2 impairments with no signal overlap 
-    - `torchsig/datasets/conf.py` -> `WidebandImpairedTrainConfig` -> `overlap_prob = 0.0`
+- Wideband with level 2 impairments
 
 ### Notes     
 - The first class of `wideband_yolo.yaml` has been modified to say `signal` because this training method is detection only. 
@@ -27,18 +25,22 @@ yolo detect train data=wideband_yolo.yaml model=yolov8x pretrained=yolov8x.pt de
 
 ## Installation with Docker
 ---
-Clone the `torchsig` repository and install using the following commands:
+The following command downloads a couple files needed to run the GNU Radio block:
 ```
-git clone https://github.com/TorchDSP/torchsig.git
-cd torchsig
-pip install .
-cd gr-spectrumdetect
-bash build_docker.sh
-bash run_docker.sh
-cd /build/gr-spectrumdetect/examples/
-source  /opt/gnuradio/v3.10/setup_env.sh
-bash trained_model_download.sh
-gnuradio-companion example.grc &
+$ cd torchsig/gr-spectrumdetect/examples
+$ bash trained_model_download.sh
+```
+Now move back to the `gr-spectrumdetect` directory, then install and run the docker container:
+```
+$ cd ..
+$ bash build_docker.sh
+$ bash run_docker.sh
+```
+Now running inside the docker:
+```
+# source  /opt/gnuradio/v3.10/setup_env.sh
+# cd /build/gr-spectrumdetect/examples/
+# gnuradio-companion example.grc &
 ```
 
 ## Installation without Docker
@@ -69,4 +71,4 @@ bash train.sh
 
 ## License
 ---
-gr-spectrumdetect is released under the MIT License. The MIT license is a popular open-source software license enabling free use, redistribution, and modifications, even for commercial purposes, provided the license is included in all copies or substantial portions of the software. TorchSig has no connection to MIT, other than through the use of this license.
+gr-spectrumdetect is released under the MIT License. The MIT license is a popular open-source software license enabling free use, redistribution, and modifications, even for commercial purposes, provided the license is included in all copies or substantial portions of the software. TorchSig has no connection to MIT, other than through the use of this license.

gr-spectrumdetect/examples/example.grc

@@ -73,7 +73,7 @@ blocks:
   id: variable
   parameters:
     comment: ''
-    value: '25000000'
+    value: '50000000'
   states:
     bus_sink: false
     bus_source: false
@@ -128,19 +128,28 @@ blocks:
   parameters:
     affinity: ''
     alias: ''
-    augment: 'False'
     center_frequency: centerFrequency
     comment: ''
-    debug_sigMF: 'False'
-    iou: '0.1'
-    max_det: '300'
+    detect_json: 'False'
+    gpu_device: '0'
+    gpu_half: 'True'
     maxoutbuf: '0'
     minoutbuf: '0'
     n_fft: nfft
     sample_rate: sampleRate
-    save: 'False'
-    trained_model: trainedModel
+    trained_narrowband_model: xcit.ckpt
+    trained_wideband_model: 11s.pt
     vector_size: vectorSize
+    wb_detect_only: 'True'
+    write_labeled_wb_images: 'False'
+    write_nb_iq_file: 'False'
+    write_wb_images: 'False'
+    write_wb_iq_file: 'False'
+    yolo_agnostic_nms: 'False'
+    yolo_augment: 'False'
+    yolo_conf: '0.25'
+    yolo_iou: '0.7'
+    yolo_max_det: '300'
   states:
     bus_sink: false
     bus_source: false
@@ -169,7 +178,7 @@ blocks:
   parameters:
     affinity: ''
     alias: ''
-    ant0: '"TX/RX"'
+    ant0: '"RX2"'
     ant1: '"RX2"'
     ant10: '"RX2"'
     ant11: '"RX2"'

gr-spectrumdetect/examples/train.sh

@@ -1,2 +1,2 @@
 #!/bin/bash
-yolo detect train data=wideband_detector_yolo.yaml model=yolov11s pretrained=yolov11s.pt device=0 epochs=1 batch=32 save=True save_period=1 single_cls=True imgsz=1024 name=11s_freeze1 cos_lr=False cache=False workers=16 freeze=1 lr0=0.0001 optimizer=SGD
+yolo detect train data=wideband_yolo.yaml model=yolov8x pretrained=yolov8x.pt device=0 epochs=1 batch=32 save=True save_period=1 single_cls=True imgsz=1024 name=8x_freeze1 cos_lr=False cache=False workers=16 freeze=1 lr0=0.0033329 optimizer=SGD

gr-spectrumdetect/examples/trained_model_download.sh

@@ -11,3 +11,25 @@ if [ $? -eq 0 ]; then
 else
     echo "Download failed."
 fi
+
+destination_path=11s.pt
+download_url=https://bucket.ltsnet.net/torchsig/models/11s.pt
+
+curl -L -o "$destination_path" "$download_url"
+
+if [ $? -eq 0 ]; then
+    echo "Download completed successfully."
+else
+    echo "Download failed."
+fi
+
+destination_path=xcit.ckpt
+download_url=https://bucket.ltsnet.net/torchsig/models/xcit.ckpt
+
+curl -L -o "$destination_path" "$download_url"
+
+if [ $? -eq 0 ]; then
+    echo "Download completed successfully."
+else
+    echo "Download failed."
+fi

gr-spectrumdetect/grc/spectrumDetect_specDetect.block.yml

@@ -5,7 +5,7 @@ category: '[spectrumDetect]'
 
 templates:
   imports: from gnuradio import spectrumDetect
-  make: spectrumDetect.specDetect(${center_frequency},${sample_rate},${vector_size}, ${n_fft}, ${trained_model}, ${augment}, ${iou}, ${save}, ${max_det}, ${debug_sigMF})
+  make: spectrumDetect.specDetect(${center_frequency},${sample_rate},${vector_size}, ${n_fft},${trained_wideband_model}, ${trained_narrowband_model}, ${yolo_augment}, ${yolo_iou}, ${yolo_conf}, ${yolo_agnostic_nms}, ${yolo_max_det}, ${write_labeled_wb_images}, ${write_wb_images}, ${write_wb_iq_file}, ${write_nb_iq_file}, ${gpu_device}, ${gpu_half}, ${detect_json}, ${wb_detect_only})
 
 parameters:
 - id: center_frequency
@@ -16,7 +16,7 @@ parameters:
 - id: sample_rate
   label: sampleRate
   dtype: float
-  default: 20000000.0
+  default: 50000000.0
 
 - id: vector_size
   label: vectorSize
@@ -28,35 +28,80 @@ parameters:
   dtype: int
   default: 1024
 
-- id: trained_model
-  label: trainedModel
+- id: trained_wideband_model
+  label: trainedWidebandModel
   dtype: string
-  default: 'detect.pt'
+  default: '11s.pt'
 
-- id: augment
-  label: yoloAugment
+- id: trained_narrowband_model
+  label: trainedNarrowbandModel
+  dtype: string
+  default: 'xcit.ckpt'
+
+- id: yolo_augment
+  label: augment
   dtype: bool
   default: False  
 
-- id: iou
-  label: yoloIOU
+- id: yolo_iou
+  label: iou
+  dtype: float
+  default: 0.7  
+
+- id: yolo_conf
+  label: conf
   dtype: float
-  default: 0.1  
+  default: 0.25   
 
-- id: save
-  label: yoloSave
+- id: yolo_agnostic_nms
+  label: agnosticNms
   dtype: bool
-  default: False    
+  default: False
 
-- id: max_det
-  label: yoloMAX_DET
+- id: yolo_max_det
+  label: maxDet
   dtype: int
   default: 300  
+
+- id: write_labeled_wb_images
+  label: writeLabeledWBImages
+  dtype: bool
+  default: False    
+
+- id: write_wb_images
+  label: writeWBImages
+  dtype: bool
+  default: False  
+
+- id: write_wb_iq_file
+  label: writeWBIQFile
+  dtype: bool
+  default: False 
 
-- id: debug_sigMF
-  label: debugOut_sigMF
+- id: write_nb_iq_file
+  label: writeNBIQFile
+  dtype: bool
+  default: False 
+
+- id: gpu_device
+  label: gpuDevice
+  dtype: string
+  default: '0' 
+
+- id: gpu_half
+  label: gpuHalf
+  dtype: bool
+  default: False 
+
+- id: detect_json
+  label: detectJson
   dtype: bool
   default: False 
+
+- id: wb_detect_only
+  label: wbDetectOnly
+  dtype: bool
+  default: True          
 
 inputs:
 - label: vector