Merge pull request #931 from anarkiwi/filt

image_inference reads from retune_fft (so gets filtered FFT stream).
IQTLabs · Oct 23, 2023 · e7861bb · e7861bb
2 parents f526ce9 + ee59a71
commit e7861bb
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 27 deletions.
diff --git a/docker/Dockerfile.base b/docker/Dockerfile.base
@@ -87,8 +87,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
   libvulkan-dev \
   python3-numpy
 WORKDIR /root
-RUN git clone https://github.com/iqtlabs/gr-iqtlabs -b 1.0.44
-RUN sed -i /SPIRV-Tools/d gr-iqtlabs/lib/CMakeLists.txt
+RUN git clone https://github.com/iqtlabs/gr-iqtlabs -b 1.0.45
 COPY --from=iqtlabs/gamutrf-vkfft:latest /root /root/gr-iqtlabs
 WORKDIR /root/gr-iqtlabs/build
 COPY --from=sigmf-builder /usr/local /usr/local

diff --git a/docs/README-airt.md b/docs/README-airt.md
@@ -85,7 +85,7 @@ install gr-iqtlabs
 $ git clone https://github.com/google/flatbuffers -b v23.5.26
 $ git clone https://github.com/nlohmann/json -b v3.11.2
 $ git clone https://github.com/deepsig/libsigmf -b v1.0.2
-$ git clone https://github.com/iqtlabs/gr-iqtlabs -b 1.0.44
+$ git clone https://github.com/iqtlabs/gr-iqtlabs -b 1.0.45
 $ mkdir -p flatbuffers/build && cd flatbuffers/build && cmake -DCMAKE_INSTALL_PREFIX=~/.conda/envs/$CONDA_DEFAULT_ENV .. && make -j $(nproc) && make install && cd ../..
 $ mkdir -p json/build && cd json/build && cmake -DCMAKE_INSTALL_PREFIX=~/.conda/envs/$CONDA_DEFAULT_ENV .. && make -j $(nproc) && make install && cd ../..
 $ mkdir -p libsigmf/build && cd libsigmf/build && cmake -DUSE_SYSTEM_JSON=ON -DUSE_SYSTEM_FLATBUFFERS=ON -DCMAKE_INSTALL_PREFIX=~/.conda/envs/$CONDA_DEFAULT_ENV -DCMAKE_CXX_FLAGS="-I $HOME/.conda/envs/$CONDA_DEFAULT_ENV/include" .. && make -j $(nproc) && make install && cd ../..

diff --git a/gamutrf/grscan.py b/gamutrf/grscan.py
@@ -151,7 +151,6 @@ def __init__(
             pretune,
         )
         self.fft_blocks = fft_blocks + self.get_db_blocks(nfft, samp_rate, scaling)
-        self.fft_to_inference_block = self.fft_blocks[-1]
 
         retune_fft = self.iqtlabs.retune_fft(
             "rx_freq",
@@ -179,7 +178,7 @@ def __init__(
         logging.info("serving FFT on %s", zmq_addr)
         self.fft_blocks.append((zeromq.pub_sink(1, 1, zmq_addr, 100, False, 65536, "")))
 
-        self.inference_blocks = []
+        self.inference_blocks = [blocks.null_sink(gr.sizeof_float * nfft)]
         if inference_output_dir:
             x = 640
             y = 640
@@ -190,26 +189,24 @@ def __init__(
             image_dir = Path(inference_output_dir, "images")
             Path(inference_output_dir).mkdir(parents=True, exist_ok=True)
             image_dir.mkdir(parents=True, exist_ok=True)
-            self.image_inference_block = self.iqtlabs.image_inference(
-                tag="rx_freq",
-                vlen=nfft,
-                x=x,
-                y=y,
-                image_dir=str(image_dir),
-                convert_alpha=255,
-                norm_alpha=0,
-                norm_beta=1,
-                norm_type=32,  # cv::NORM_MINMAX = 32
-                colormap=16,  # cv::COLORMAP_VIRIDIS = 16, cv::COLORMAP_TURBO = 20,
-                interpolation=1,  # cv::INTER_LINEAR = 1,
-                flip=0,
-                min_peak_points=inference_min_db,
-                model_server=inference_model_server,
-                model_name=inference_model_name,
-            )
             self.inference_blocks = [
-                blocks.stream_to_vector(gr.sizeof_float * nfft, 1),
-                self.image_inference_block,
+                self.iqtlabs.image_inference(
+                    tag="rx_freq",
+                    vlen=nfft,
+                    x=x,
+                    y=y,
+                    image_dir=str(image_dir),
+                    convert_alpha=255,
+                    norm_alpha=0,
+                    norm_beta=1,
+                    norm_type=32,  # cv::NORM_MINMAX = 32
+                    colormap=16,  # cv::COLORMAP_VIRIDIS = 16, cv::COLORMAP_TURBO = 20,
+                    interpolation=1,  # cv::INTER_LINEAR = 1,
+                    flip=0,
+                    min_peak_points=inference_min_db,
+                    model_server=inference_model_server,
+                    model_name=inference_model_name,
+                ),
                 yolo_bbox(
                     str(Path(inference_output_dir, "predictions")),
                     inference_min_confidence,
@@ -223,17 +220,18 @@ def __init__(
         else:
             self.msg_connect((retune_fft, "tune"), (self.sources[0], cmd_port))
         self.connect_blocks(self.sources[0], self.sources[1:])
-        self.connect_blocks(self.fft_to_inference_block, self.inference_blocks)
+        self.connect((retune_fft, 1), (self.inference_blocks[0], 0))
+        self.connect_blocks(self.inference_blocks[0], self.inference_blocks[1:])
         for pipeline_blocks in (
             self.fft_blocks,
             self.samples_blocks,
         ):
             self.connect_blocks(self.sources[-1], pipeline_blocks)
 
-    def connect_blocks(self, source, other_blocks):
+    def connect_blocks(self, source, other_blocks, last_block_port=0):
         last_block = source
         for block in other_blocks:
-            self.connect((last_block, 0), (block, 0))
+            self.connect((last_block, last_block_port), (block, 0))
             last_block = block
 
     def get_db_blocks(self, nfft, samp_rate, scaling):