mediapipe-rs/mediapipe/graphs/template_matching/template_matching_desktop.pbtxt
Victor Dudochkin 5578aa50e8 code fill
2022-03-01 19:04:01 +07:00

142 lines
3.9 KiB
Plaintext

# MediaPipe graph that performs object detection on desktop with TensorFlow Lite
# on CPU.
# Used in the example in
# mediapipe/examples/desktop/template_matching:template_matching_tflite
# max_queue_size limits the number of packets enqueued on any input stream
# by throttling inputs to the graph. This makes the graph only process one
# frame per time.
max_queue_size: 1
# Decodes an input video file into images and a video header.
node {
calculator: "OpenCvVideoDecoderCalculator"
input_side_packet: "INPUT_FILE_PATH:input_video_path"
output_stream: "VIDEO:input_video"
output_stream: "VIDEO_PRESTREAM:input_video_header"
}
node: {
calculator: "ImageTransformationCalculator"
input_stream: "IMAGE:input_video"
output_stream: "IMAGE:scaled_input_video"
node_options: {
[type.googleapis.com/mediapipe.ImageTransformationCalculatorOptions] {
output_width: 640
output_height: 640
scale_mode: FILL_AND_CROP
}
}
}
node {
calculator: "ImagePropertiesCalculator"
input_stream: "IMAGE:scaled_input_video"
output_stream: "SIZE:input_video_size"
}
node {
calculator: "FeatureDetectorCalculator"
input_stream: "IMAGE:scaled_input_video"
output_stream: "FEATURES:features"
output_stream: "LANDMARKS:landmarks"
output_stream: "PATCHES:patches"
}
# input tensors: 200*32*32*1 float
# output tensors: 200*40 float, only first keypoint.size()*40 is knift features,
# rest is padded by zero.
node {
calculator: "TfLiteInferenceCalculator"
input_stream: "TENSORS:patches"
output_stream: "TENSORS:knift_feature_tensors"
node_options: {
[type.googleapis.com/mediapipe.TfLiteInferenceCalculatorOptions] {
model_path: "mediapipe/models/knift_float.tflite"
}
}
}
node {
calculator: "TfLiteTensorsToFloatsCalculator"
input_stream: "TENSORS:knift_feature_tensors"
output_stream: "FLOATS:knift_feature_floats"
}
node {
calculator: "BoxDetectorCalculator"
input_stream: "FEATURES:features"
input_stream: "IMAGE_SIZE:input_video_size"
input_stream: "DESCRIPTORS:knift_feature_floats"
output_stream: "BOXES:detections"
node_options: {
[type.googleapis.com/mediapipe.BoxDetectorCalculatorOptions] {
detector_options {
index_type: OPENCV_BF
detect_every_n_frame: 1
}
index_proto_filename: "mediapipe/models/knift_index.pb"
}
}
}
node {
calculator: "TimedBoxListIdToLabelCalculator"
input_stream: "detections"
output_stream: "labeled_detections"
node_options: {
[type.googleapis.com/mediapipe.TimedBoxListIdToLabelCalculatorOptions] {
label_map_path: "mediapipe/models/knift_labelmap.txt"
}
}
}
node {
calculator: "TimedBoxListToRenderDataCalculator"
input_stream: "BOX_LIST:labeled_detections"
output_stream: "RENDER_DATA:box_render_data"
node_options: {
[type.googleapis.com/mediapipe.TimedBoxListToRenderDataCalculatorOptions] {
box_color { r: 255 g: 0 b: 0 }
thickness: 5.0
}
}
}
node {
calculator: "LandmarksToRenderDataCalculator"
input_stream: "NORM_LANDMARKS:landmarks"
output_stream: "RENDER_DATA:landmarks_render_data"
node_options: {
[type.googleapis.com/mediapipe.LandmarksToRenderDataCalculatorOptions] {
landmark_color { r: 0 g: 255 b: 0 }
thickness: 2.0
}
}
}
# Draws annotations and overlays them on top of the input images.
node {
calculator: "AnnotationOverlayCalculator"
input_stream: "IMAGE:input_video"
input_stream: "box_render_data"
input_stream: "landmarks_render_data"
output_stream: "IMAGE:output_video"
}
# Encodes the annotated images into a video file, adopting properties specified
# in the input video header, e.g., video framerate.
node {
calculator: "OpenCvVideoEncoderCalculator"
input_stream: "VIDEO:output_video"
input_stream: "VIDEO_PRESTREAM:input_video_header"
input_side_packet: "OUTPUT_FILE_PATH:output_video_path"
node_options: {
[type.googleapis.com/mediapipe.OpenCvVideoEncoderCalculatorOptions]: {
codec: "avc1"
video_format: "mp4"
}
}
}