tests: Add hailo and imx500 tests

tests/hailo.py

@@ -0,0 +1,95 @@
+#!/bin/python3
+
+import os
+
+try:
+    from picamera2.devices import Hailo, hailo_architecture
+except ImportError:
+    print("SKIPPED (hailo_platform not installed)")
+    quit()
+
+from picamera2 import Picamera2
+
+# Check a Hailo device is present.
+arch = hailo_architecture()
+if arch is None:
+    print("SKIPPED (no Hailo device found)")
+    quit()
+
+# Pick models based on architecture.
+if arch == 'HAILO10H':
+    detect_model = '/usr/share/hailo-models/yolov8m_h10.hef'
+    pose_model = '/usr/share/hailo-models/yolov8s_pose_h10.hef'
+else:
+    detect_model = '/usr/share/hailo-models/yolov8s_h8l.hef'
+    pose_model = '/usr/share/hailo-models/yolov8s_pose_h8l_pi.hef'
+
+if not os.path.exists(detect_model):
+    print("SKIPPED (detection model not found:", detect_model + ")")
+    quit()
+if not os.path.exists(pose_model):
+    print("SKIPPED (pose model not found:", pose_model + ")")
+    quit()
+
+NUM_FRAMES = 30
+
+# Test detection model over 30 frames.
+print("Testing detection model:", detect_model)
+with Hailo(detect_model) as hailo:
+    input_shape = hailo.get_input_shape()
+    print("Input shape:", input_shape)
+    if len(input_shape) != 3:
+        print("ERROR: expected 3-dimensional input shape, got", len(input_shape))
+
+    inputs, outputs = hailo.describe()
+    print("Model inputs:", len(inputs), "outputs:", len(outputs))
+    if len(inputs) < 1:
+        print("ERROR: expected at least 1 input layer")
+    if len(outputs) < 1:
+        print("ERROR: expected at least 1 output layer")
+
+    with Picamera2() as picam2:
+        model_h, model_w = input_shape[0], input_shape[1]
+        config = picam2.create_preview_configuration(
+            main={'size': (1920, 1080), 'format': 'XRGB8888'},
+            lores={'size': (model_w, model_h), 'format': 'RGB888'}
+        )
+        picam2.configure(config)
+        picam2.start()
+
+        for i in range(NUM_FRAMES):
+            frame = picam2.capture_array('lores')
+            result = hailo.run(frame)
+            if result is None:
+                print("ERROR: detection inference returned None on frame", i)
+                break
+        else:
+            print("Detection model: all", NUM_FRAMES, "frames returned results")
+
+        picam2.stop()
+
+# Test pose estimation model over 30 frames.
+print("Testing pose model:", pose_model)
+with Hailo(pose_model) as hailo:
+    input_shape = hailo.get_input_shape()
+    print("Input shape:", input_shape)
+
+    with Picamera2() as picam2:
+        model_h, model_w = input_shape[0], input_shape[1]
+        config = picam2.create_preview_configuration(
+            main={'size': (1920, 1080), 'format': 'XRGB8888'},
+            lores={'size': (model_w, model_h), 'format': 'RGB888'}
+        )
+        picam2.configure(config)
+        picam2.start()
+
+        for i in range(NUM_FRAMES):
+            frame = picam2.capture_array('lores')
+            result = hailo.run(frame)
+            if result is None:
+                print("ERROR: pose inference returned None on frame", i)
+                break
+        else:
+            print("Pose model: all", NUM_FRAMES, "frames returned results")
+
+        picam2.stop()

tests/imx500.py

@@ -0,0 +1,49 @@
+#!/bin/python3
+
+import os
+
+from picamera2 import Picamera2
+from picamera2.devices.imx500 import IMX500
+
+# Check if an imx500 camera is connected.
+camera_info = Picamera2.global_camera_info()
+camera_num = next((c['Num'] for c in camera_info if c['Model'] == 'imx500'), None)
+
+if camera_num is None:
+    print("SKIPPED (no imx500 camera found)")
+    quit()
+
+model_path = '/usr/share/imx500-models/imx500_network_ssd_mobilenetv2_fpnlite_320x320_pp.rpk'
+if not os.path.exists(model_path):
+    print("SKIPPED (model file not found:", model_path + ")")
+    quit()
+
+imx500 = IMX500(model_path)
+
+# Test getting the device ID.
+device_id = imx500.get_device_id()
+print("Device ID:", device_id)
+if not device_id:
+    print("ERROR: empty device ID")
+
+# Start the camera and capture some frames with output tensors.
+picam2 = Picamera2(imx500.camera_num)
+config = picam2.create_preview_configuration(buffer_count=12)
+picam2.configure(config)
+picam2.start()
+
+NUM_FRAMES = 30
+
+tensor_count = 0
+for _ in range(NUM_FRAMES):
+    metadata = picam2.capture_metadata()
+    outputs = imx500.get_outputs(metadata, add_batch=True)
+    if outputs is not None and len(outputs) > 0:
+        tensor_count += 1
+
+print("Got output tensors on", tensor_count, "of", NUM_FRAMES, "frames")
+if tensor_count < NUM_FRAMES // 2:
+    print("ERROR: expected at least", NUM_FRAMES // 2, "frames with output tensors")
+
+picam2.stop()
+picam2.close()

tests/test_list.txt

@@ -85,6 +85,8 @@ tests/encoder_import.py
 tests/encoder_start_stop.py
 tests/ffmpeg_abort.py
 tests/grey_world.py
+tests/hailo.py
+tests/imx500.py
 tests/imx708_device.py
 tests/large_datagram.py
 tests/mjpeg_server.py