error message:
Traceback (most recent call last):
File "/home/prakrisht/-------------/detect_from_webcam.py", line 173, in <module>
run_inference(detection_model, category_index, cap)
File "/home/prakrisht/----------------/detect_from_webcam.py", line 60, in run_inference
output_dict = run_inference_for_single_image(model, image_np)
File "/home/prakrisht/-----------------/detect_from_webcam.py", line 31, in run_inference_for_single_image
output_dict = model(input_tensor)
File "/home/prakrisht/anaconda3/envs/me_test/lib/python3.9/site-packages/tensorflow/python/saved_model/load.py", line 816, in _call_attribute
return instance.__call__(*args, **kwargs)
File "/home/prakrisht/anaconda3/envs/me_test/lib/python3.9/site-packages/tensorflow/python/util/traceback_utils.py", line 153, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/home/prakrisht/anaconda3/envs/me_test/lib/python3.9/site-packages/tensorflow/python/eager/polymorphic_function/function_type_utils.py", line 446, in bind_function_inputs
raise TypeError(
TypeError: Binding inputs to tf.function failed due to `Can not cast TensorSpec(shape=(1,), dtype=tf.float32, name='input_tensor') to TensorSpec(shape=(1, None, None, 3), dtype=tf.uint8, name='input_tensor')`. Received args: (<tf.Tensor: shape=(1,), dtype=float32, numpy=array([nan], dtype=float32)>,) and kwargs: {} for signature: (input_tensor: TensorSpec(shape=(1, None, None, 3), dtype=tf.uint8, name='input_tensor')).
Code for the detection: summary:
- we run inference on the camera feed
- using a resnet101 model trained on a custom dataset
- the basic code works perfectly with images
- it was working perfectly some time ago with webcam as well, but doesn't work anymore.
import numpy as np
import argparse
import tensorflow as tf
import cv2
from object_detection.utils import ops as utils_ops
from object_detection.utils import label_map_util
from object_detection.utils import visualization_utils as vis_util
# patch tf1 into `utils.ops`
utils_ops.tf = tf.compat.v1
# Patch the location of gfile
tf.gfile = tf.io.gfile
def load_model(model_path):
model = tf.saved_model.load(model_path)
return model
def run_inference_for_single_image(model, image):
image = np.asarray(image).astype(np.float32)
# The input needs to be a tensor, convert it using `tf.convert_to_tensor`.
input_tensor = tf.convert_to_tensor(image)
# The model expects a batch of images, so add an axis with `tf.newaxis`.
input_tensor = input_tensor[tf.newaxis,...]
# Run inference
output_dict = model(input_tensor)
# All outputs are batches tensors.
# Convert to numpy arrays, and take index [0] to remove the batch dimension.
# We're only interested in the first num_detections.
num_detections = int(output_dict.pop('num_detections'))
output_dict = {key: value[0, :num_detections].numpy()
for key, value in output_dict.items()}
output_dict['num_detections'] = num_detections
# detection_classes should be ints.
output_dict['detection_classes'] = output_dict['detection_classes'].astype(np.int64)
# Handle models with masks:
if 'detection_masks' in output_dict:
# Reframe the the bbox mask to the image size.
detection_masks_reframed = utils_ops.reframe_box_masks_to_image_masks(
output_dict['detection_masks'], output_dict['detection_boxes'],
image.shape[0], image.shape[1])
detection_masks_reframed = tf.cast(detection_masks_reframed > 0.5, tf.uint8)
output_dict['detection_masks_reframed'] = detection_masks_reframed.numpy()
return output_dict
def run_inference(model, category_index, cap):
while True:
ret, image_np = cap.read()
# Actual detection.
output_dict = run_inference_for_single_image(model, image_np)
# Visualization of the results of a detection.
#print(output_dict)
draw_boxes_on_image(
image_np,
output_dict['detection_boxes'],
output_dict['detection_classes'],
output_dict['detection_scores'],
category_index,
#instance_masks=output_dict.get('detection_masks_reframed', None),
#use_normalized_coordinates=True,
line_thickness=8,
font_scale=1)
cv2.imshow('object_detection', cv2.resize(image_np, (800, 600)))
if cv2.waitKey(25) & 0xFF == ord('q'):
cap.release()
cv2.destroyAllWindows()
break
import cv2
import numpy as np
def draw_boxes_on_image(image_np, boxes, classes, scores, category_index,
threshold=0.5, line_thickness=8, font=cv2.FONT_HERSHEY_SIMPLEX,
font_scale=0.7, font_thickness=2):
"""
Draw bounding boxes on the image with class-specific colors, showing labels with the highest score on top.
Alternate label placement between the left and right sides of the bounding box.
Args:
- image_np: Numpy array of the image.
- boxes: Array of bounding boxes, with shape [N, 4].
- classes: Array of class indices corresponding to the boxes.
- scores: Array of confidence scores corresponding to the boxes.
- category_index: Dictionary mapping class indices to class names.
- threshold: Confidence threshold for displaying the bounding box.
- line_thickness: Thickness of the bounding box lines.
- font: Font type for the label.
- font_scale: Scale (size) of the font for the label.
- font_thickness: Thickness of the font for the label.
"""
# Assign a unique color for each class
colors = {}
np.random.seed(42) # For reproducibility
for class_id in np.unique(classes):
colors[class_id] = tuple(np.random.randint(0, 256, 3).tolist())
# Sort detections by score in descending order
sorted_indices = np.argsort(-scores)
for i in sorted_indices:
if scores[i] > threshold:
box = boxes[i]
class_id = int(classes[i])
score = scores[i]
ymin, xmin, ymax, xmax = box
im_height, im_width, _ = image_np.shape
(left, right, top, bottom) = (int(xmin * im_width), int(xmax * im_width),
int(ymin * im_height), int(ymax * im_height))
# Draw bounding box
color = colors[class_id]
cv2.rectangle(image_np, (left, top), (right, bottom), color, line_thickness)
# Prepare label text
#strip "rotation" from the label
label_text = category_index[class_id]["name"]
if label_text.find("rotation") != -1:
label_text = label_text[:label_text.find("rotation")]
label = f'{label_text}: {int(score * 100)}%'
label_size, base_line = cv2.getTextSize(label, font, font_scale, font_thickness)
top = max(top, label_size[1])
# Determine label position (alternate between left and right)
if i % 2 == 0:
# Left side of the box
label_left = left
else:
# Right side of the box
label_left = right - label_size[0]
# Draw label background and text
cv2.rectangle(image_np, (label_left, top - label_size[1]),
(label_left + label_size[0], top + base_line), color, cv2.FILLED)
cv2.putText(image_np, label, (label_left, top), font, font_scale, (0, 0, 0), font_thickness)
# Adjust the top position to avoid overlapping with the next label
top -= (label_size[1] + base_line + 5)
return image_np
if __name__ == '__main__':
model_path = './inference_graph_resnet101/saved_model'
label_map_path = './-------------/1_label_map.pbtxt'
detection_model = load_model(model_path)
category_index = label_map_util.create_category_index_from_labelmap(label_map_path, use_display_name=True)
cap = cv2.VideoCapture(-1)
#check camera and display the feed
#print all available cameras
for i in range(0, 10):
cap = cv2.VideoCapture(i)
if not cap.isOpened():
print(f"Camera {i} is not available")
else:
print(f"Camera {i} is available")
cap.release()
run_inference(detection_model, category_index, cap)
Result for camera test:
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video0): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video0): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 0 is not available
Camera 1 is available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video2): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 2 is not available
Camera 3 is available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video4): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 4 is not available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video5): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 5 is not available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video6): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 6 is not available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video7): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 7 is not available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video8): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 8 is not available
[ WARN:[email protected]] global cap_v4l.cpp:999 open VIDEOIO(V4L2:/dev/video9): can't open camera by index
[ERROR:[email protected]] global obsensor_uvc_stream_channel.cpp:158 getStreamChannelGroup Camera index out of range
Camera 9 is not available
Troubleshooting done already:
- tried different camera indices.
- tested with set of images, it works perfectly
- checked model paths etc.
no idea, where to head next....
