mirror of
https://github.com/azaion/detections.git
synced 2026-04-22 11:06:32 +00:00
fix some cython code
This commit is contained in:
Oleksandr Bezdieniezhnykh
+43
-26
@@ -8,28 +8,42 @@ cimport constants_inf
|
||||
from ai_availability_status cimport AIAvailabilityEnum, AIAvailabilityStatus
|
||||
from annotation cimport Detection, Annotation
|
||||
from ai_config cimport AIRecognitionConfig
|
||||
from engines.inference_engine cimport InferenceEngine
|
||||
from loader_http_client cimport LoaderHttpClient
|
||||
from threading import Thread
|
||||
from engines import tensor_gpu_index, create_engine
|
||||
if tensor_gpu_index > -1:
|
||||
from engines.tensorrt_engine import TensorRTEngine
|
||||
|
||||
|
||||
from engines import EngineClass
|
||||
|
||||
|
||||
cdef class Inference:
|
||||
cdef LoaderHttpClient loader_client
|
||||
cdef InferenceEngine engine
|
||||
cdef object _annotation_callback
|
||||
cdef object _status_callback
|
||||
cdef Annotation _previous_annotation
|
||||
cdef dict[str, list[Detection]] _tile_detections
|
||||
cdef dict[str, int] detection_counts
|
||||
cdef AIRecognitionConfig ai_config
|
||||
cdef bint stop_signal
|
||||
cdef public AIAvailabilityStatus ai_availability_status
|
||||
cdef str model_input
|
||||
cdef int model_width
|
||||
cdef int model_height
|
||||
cdef bytes _converted_model_bytes
|
||||
cdef bint is_building_engine
|
||||
|
||||
def __init__(self, loader_client):
|
||||
self.loader_client = loader_client
|
||||
self._annotation_callback = None
|
||||
self._status_callback = None
|
||||
self.stop_signal = False
|
||||
self.model_input = None
|
||||
self.stop_signal = <bint>False
|
||||
self.model_input = <str>None
|
||||
self.model_width = 0
|
||||
self.model_height = 0
|
||||
self.detection_counts = {}
|
||||
self.engine = None
|
||||
self.is_building_engine = False
|
||||
self.engine = <InferenceEngine>None
|
||||
self.is_building_engine = <bint>False
|
||||
self.ai_availability_status = AIAvailabilityStatus()
|
||||
self._converted_model_bytes = None
|
||||
self._converted_model_bytes = <bytes>None
|
||||
self.init_ai()
|
||||
|
||||
@property
|
||||
@@ -49,7 +63,7 @@ cdef class Inference:
|
||||
try:
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.CONVERTING)
|
||||
models_dir = constants_inf.MODELS_FOLDER
|
||||
model_bytes = TensorRTEngine.convert_from_onnx(onnx_engine_bytes)
|
||||
model_bytes = EngineClass.convert_from_onnx(onnx_engine_bytes)
|
||||
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.UPLOADING)
|
||||
res = self.loader_client.upload_big_small_resource(model_bytes, engine_filename, models_dir)
|
||||
@@ -60,9 +74,9 @@ cdef class Inference:
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.ENABLED)
|
||||
except Exception as e:
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.ERROR, <str> str(e))
|
||||
self._converted_model_bytes = None
|
||||
self._converted_model_bytes = <bytes>None
|
||||
finally:
|
||||
self.is_building_engine = False
|
||||
self.is_building_engine = <bint>False
|
||||
|
||||
cdef init_ai(self):
|
||||
constants_inf.log(<str> 'init AI...')
|
||||
@@ -74,24 +88,24 @@ cdef class Inference:
|
||||
|
||||
if self._converted_model_bytes is not None:
|
||||
try:
|
||||
self.engine = TensorRTEngine(self._converted_model_bytes)
|
||||
self.engine = EngineClass(self._converted_model_bytes)
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.ENABLED)
|
||||
self.model_height, self.model_width = self.engine.get_input_shape()
|
||||
except Exception as e:
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.ERROR, <str> str(e))
|
||||
finally:
|
||||
self._converted_model_bytes = None # Consume the bytes
|
||||
self._converted_model_bytes = <bytes>None
|
||||
return
|
||||
|
||||
models_dir = constants_inf.MODELS_FOLDER
|
||||
if tensor_gpu_index > -1:
|
||||
engine_filename = EngineClass.get_engine_filename()
|
||||
if engine_filename is not None:
|
||||
try:
|
||||
engine_filename = TensorRTEngine.get_engine_filename(0)
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.DOWNLOADING)
|
||||
res = self.loader_client.load_big_small_resource(engine_filename, models_dir)
|
||||
if res.err is not None:
|
||||
raise Exception(res.err)
|
||||
self.engine = TensorRTEngine(res.data)
|
||||
self.engine = EngineClass(res.data)
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.ENABLED)
|
||||
except Exception as e:
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.WARNING, <str>str(e))
|
||||
@@ -103,7 +117,8 @@ cdef class Inference:
|
||||
thread.start()
|
||||
return
|
||||
else:
|
||||
self.engine = create_engine(<bytes>self.get_onnx_engine_bytes())
|
||||
self.engine = EngineClass(<bytes>self.get_onnx_engine_bytes())
|
||||
self.ai_availability_status.set_status(AIAvailabilityEnum.ENABLED)
|
||||
self.is_building_engine = False
|
||||
|
||||
self.model_height, self.model_width = self.engine.get_input_shape()
|
||||
@@ -178,7 +193,7 @@ cdef class Inference:
|
||||
|
||||
cdef bint is_video(self, str filepath):
|
||||
mime_type, _ = mimetypes.guess_type(<str>filepath)
|
||||
return mime_type and mime_type.startswith("video")
|
||||
return <bint>(mime_type and mime_type.startswith("video"))
|
||||
|
||||
cdef split_list_extend(self, lst, chunk_size):
|
||||
chunks = [lst[i:i + chunk_size] for i in range(0, len(lst), chunk_size)]
|
||||
@@ -200,7 +215,7 @@ cdef class Inference:
|
||||
|
||||
self._annotation_callback = annotation_callback
|
||||
self._status_callback = status_callback
|
||||
self.stop_signal = False
|
||||
self.stop_signal = <bint>False
|
||||
self.init_ai()
|
||||
if self.engine is None:
|
||||
constants_inf.log(<str> "AI engine not available. Conversion may be in progress. Skipping inference.")
|
||||
@@ -246,9 +261,9 @@ cdef class Inference:
|
||||
cdef int frame_count = 0
|
||||
cdef int batch_count = 0
|
||||
cdef list batch_frames = []
|
||||
cdef list[int] batch_timestamps = []
|
||||
cdef list[long] batch_timestamps = []
|
||||
cdef Annotation annotation
|
||||
self._previous_annotation = None
|
||||
self._previous_annotation = <Annotation>None
|
||||
|
||||
v_input = cv2.VideoCapture(<str>video_name)
|
||||
if not v_input.isOpened():
|
||||
@@ -267,7 +282,7 @@ cdef class Inference:
|
||||
frame_count += 1
|
||||
if frame_count % ai_config.frame_period_recognition == 0:
|
||||
batch_frames.append(frame)
|
||||
batch_timestamps.append(int(v_input.get(cv2.CAP_PROP_POS_MSEC)))
|
||||
batch_timestamps.append(<long>v_input.get(cv2.CAP_PROP_POS_MSEC))
|
||||
|
||||
if len(batch_frames) == self.engine.get_batch_size():
|
||||
batch_count += 1
|
||||
@@ -308,7 +323,8 @@ cdef class Inference:
|
||||
self.detection_counts[annotation.original_media_name] = self.detection_counts.get(annotation.original_media_name, 0) + 1
|
||||
if self._annotation_callback is not None:
|
||||
percent = int(frame_count * 100 / total_frames) if total_frames > 0 else 0
|
||||
self._annotation_callback(annotation, percent)
|
||||
cb = self._annotation_callback
|
||||
cb(annotation, percent)
|
||||
|
||||
cdef _process_images(self, AIRecognitionConfig ai_config, list[str] image_paths):
|
||||
cdef list frame_data
|
||||
@@ -343,8 +359,9 @@ cdef class Inference:
|
||||
|
||||
cdef send_detection_status(self):
|
||||
if self._status_callback is not None:
|
||||
cb = self._status_callback
|
||||
for media_name in self.detection_counts.keys():
|
||||
self._status_callback(media_name, self.detection_counts[media_name])
|
||||
cb(media_name, self.detection_counts[media_name])
|
||||
self.detection_counts.clear()
|
||||
|
||||
cdef split_to_tiles(self, frame, path, tile_size, overlap_percent):
|
||||
|
||||
Reference in New Issue
Block a user