Add a script to plot hot pixels

README.md

@@ -164,7 +164,7 @@ class UsbConfiguration:
 
 See _python/tests_ for different usage examples. `python/tests/display.py` implements a live event viewer with exponential decays caculated by the GPU. It requires vispy and glfw (`pip install vispy glfw`).
 
-Some tests generate plots and require Pillow and Plotly (`pip install pillow plotly`).
+Some tests generate plots and require Plotly (`pip install plotly pandas kaleido`).
 
 ## Contribute
 

python/.gitignore

@@ -3,3 +3,4 @@ tests/*.raw
 *.so
 *.pyd
 __pycache__
+tests/hot_pixels

python/tests/find_hot_pixels.py

@@ -0,0 +1,95 @@
+import pathlib
+
+import neuromorphic_drivers as nd
+import numpy as np
+import plotly.express
+
+nd.print_device_list()
+
+output = pathlib.Path(__file__).resolve().parent / "hot_pixels"
+output.mkdir(exist_ok=True)
+
+diff_off_to_counts: dict[int, np.ndarray] = {}
+diff_off_to_duration: dict[int, float] = {}
+
+with nd.open(configuration=nd.prophesee_evk4.Configuration()) as device:
+    diff_off = 60
+    start_t = 0
+    next_t = 1000000
+    state = 0
+    total = 0
+    counts = np.zeros(
+        (device.properties().width, device.properties().height), dtype=np.uint64
+    )
+
+    print("wait for 1 seconds")
+    for status, packet in device:
+        if diff_off == 45:
+            break
+        if "dvs_events" in packet:
+            if state == 1:
+                events = packet["dvs_events"][packet["dvs_events"]["t"] >= start_t]
+                total += len(events)
+                np.add.at(counts, (events["x"], events["y"]), 1)
+            if packet["dvs_events"]["t"][-1] > next_t:
+                match state:
+                    case 0:
+                        print("start counting")
+                        start_t = packet["dvs_events"]["t"][-1] + 1
+                        next_t += 1000000
+                        total = 0
+                        counts.fill(0)
+                        state = 1
+                    case 1:
+                        duration = packet["dvs_events"]["t"][-1] - start_t
+                        print(
+                            f"{diff_off=}, {total=}, {duration=}, {total / duration * 1e6} events/s"
+                        )
+                        diff_off_to_counts[diff_off] = counts
+                        diff_off_to_duration[diff_off] = float(duration)
+                        counts = np.zeros(
+                            (device.properties().width, device.properties().height),
+                            dtype=np.uint64,
+                        )
+                        next_t += 1000000
+                        state = 0
+                        diff_off -= 1
+                        device.update_configuration(
+                            nd.prophesee_evk4.Configuration(
+                                biases=nd.prophesee_evk4.Biases(
+                                    diff_off=diff_off,
+                                )
+                            )
+                        )
+                        print("wait for 1 seconds")
+
+
+for diff_off, counts in diff_off_to_counts.items():
+    log_counts = np.log((counts + 1).astype(np.float64)).transpose()
+    figure = plotly.express.imshow(log_counts, color_continuous_scale="thermal")
+    figure.update_layout(
+        width=1280,
+        height=720,
+        margin=dict(l=0, r=0, b=0, t=0),
+        xaxis={"visible": False, "showticklabels": False},
+        yaxis={"visible": False, "showticklabels": False},
+        showlegend=False,
+    )
+    figure.update_coloraxes(showscale=False)
+    figure.write_image(output / f"diff_off_{diff_off}_map.png")
+    sorted_counts = np.sort(counts[counts > 0])[::-1].astype(np.float64) * (
+        1e6 / diff_off_to_duration[diff_off],
+    )
+    print(f"{diff_off=}, maximum={sorted_counts[0]} events/s")
+    figure = plotly.express.line(
+        x=np.arange(1, len(sorted_counts) + 1),
+        y=sorted_counts,
+        log_y=True,
+        log_x=True,
+    )
+    figure.update_layout(
+        width=1280,
+        height=720,
+        showlegend=False,
+    )
+    figure.write_image(output / f"diff_off_{diff_off}_count.png")