michael/scrypted
1
0
Fork 0
mirror of https://github.com/koush/scrypted.git synced 2026-05-06 14:10:28 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

tensorflow-lite: use dict vs queue for perf

Browse Source
This commit is contained in:
Koushik Dutta
2024-05-07 10:52:06 -07:00
parent 540b990a08
commit b4e6821da8
3 changed files with 56 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
plugins/tensorflow-lite/package-lock.json generated
View File

@@ -1,12 +1,12 @@
{
"name": "@scrypted/tensorflow-lite",
"version": "0.1.59",
"version": "0.1.60",
"lockfileVersion": 2,
"requires": true,
"packages": {
"": {
"name": "@scrypted/tensorflow-lite",
"version": "0.1.59",
"version": "0.1.60",
"devDependencies": {
"@scrypted/sdk": "file:../../sdk"
}

2
plugins/tensorflow-lite/package.json
View File

@@ -53,5 +53,5 @@
"devDependencies": {
"@scrypted/sdk": "file:../../sdk"
},
"version": "0.1.59"
"version": "0.1.60"
}

99
plugins/tensorflow-lite/src/tflite/__init__.py
View File

@@ -1,6 +1,7 @@
from __future__ import annotations
import json
import threading
from PIL import Image
from pycoral.adapters import detect
@@ -121,7 +122,8 @@ class TensorFlowLitePlugin(
labels_contents = open(labelsFile, "r").read()
self.labels = parse_label_contents(labels_contents)
self.interpreters = queue.Queue()
self.interpreters = {}
available_interpreters = []
self.interpreter_count = 0
def downloadModel():
@@ -145,7 +147,7 @@ class TensorFlowLitePlugin(
"shape"
]
self.input_details = int(width), int(height), int(channels)
self.interpreters.put(interpreter)
available_interpreters.append(interpreter)
self.interpreter_count = self.interpreter_count + 1
print("added tpu %s" % (edge_tpu))
except Exception as e:
@@ -165,12 +167,20 @@ class TensorFlowLitePlugin(
interpreter.allocate_tensors()
_, height, width, channels = interpreter.get_input_details()[0]["shape"]
self.input_details = int(width), int(height), int(channels)
self.interpreters.put(interpreter)
available_interpreters.append(interpreter)
self.interpreter_count = self.interpreter_count + 1
print(modelFile, labelsFile)
def executor_initializer():
thread_name = threading.current_thread().name
interpreter = available_interpreters.pop()
self.interpreters[thread_name] = interpreter
print('Interpreter initialized on thread {}'.format(thread_name))
self.executor = concurrent.futures.ThreadPoolExecutor(
initializer=executor_initializer,
max_workers=self.interpreter_count,
thread_name_prefix="tflite",
)
@@ -208,53 +218,50 @@ class TensorFlowLitePlugin(
async def detect_once(self, input: Image.Image, settings: Any, src_size, cvss):
def predict():
interpreter = self.interpreters.get()
try:
if self.yolo:
tensor_index = input_details(interpreter, "index")
interpreter = self.interpreters[threading.current_thread().name]
if self.yolo:
tensor_index = input_details(interpreter, "index")
im = np.stack([input])
i = interpreter.get_input_details()[0]
if i["dtype"] == np.int8:
scale, zero_point = i["quantization"]
if scale == 0.003986024297773838 and zero_point == -128:
# fast path for quantization 1/255 = 0.003986024297773838
im = im.view(np.int8)
im -= 128
else:
im = im.astype(np.float32) / (255.0 * scale)
im = (im + zero_point).astype(np.int8) # de-scale
im = np.stack([input])
i = interpreter.get_input_details()[0]
if i["dtype"] == np.int8:
scale, zero_point = i["quantization"]
if scale == 0.003986024297773838 and zero_point == -128:
# fast path for quantization 1/255 = 0.003986024297773838
im = im.view(np.int8)
im -= 128
else:
# this code path is unused.
im = im.astype(np.float32) / 255.0
interpreter.set_tensor(tensor_index, im)
interpreter.invoke()
output_details = interpreter.get_output_details()
output = output_details[0]
x = interpreter.get_tensor(output["index"])
input_scale = self.get_input_details()[0]
if x.dtype == np.int8:
scale, zero_point = output["quantization"]
threshold = yolo.defaultThreshold / scale + zero_point
combined_scale = scale * input_scale
objs = yolo.parse_yolov9(
x[0],
threshold,
scale=lambda v: (v - zero_point) * combined_scale,
confidence_scale=lambda v: (v - zero_point) * scale,
)
else:
# this code path is unused.
objs = yolo.parse_yolov9(x[0], scale=lambda v: v * input_scale)
im = im.astype(np.float32) / (255.0 * scale)
im = (im + zero_point).astype(np.int8) # de-scale
else:
tflite_common.set_input(interpreter, input)
interpreter.invoke()
objs = detect.get_objects(
interpreter, score_threshold=0.2, image_scale=(1, 1)
# this code path is unused.
im = im.astype(np.float32) / 255.0
interpreter.set_tensor(tensor_index, im)
interpreter.invoke()
output_details = interpreter.get_output_details()
output = output_details[0]
x = interpreter.get_tensor(output["index"])
input_scale = self.get_input_details()[0]
if x.dtype == np.int8:
scale, zero_point = output["quantization"]
threshold = yolo.defaultThreshold / scale + zero_point
combined_scale = scale * input_scale
objs = yolo.parse_yolov9(
x[0],
threshold,
scale=lambda v: (v - zero_point) * combined_scale,
confidence_scale=lambda v: (v - zero_point) * scale,
)
return objs
finally:
self.interpreters.put(interpreter)
else:
# this code path is unused.
objs = yolo.parse_yolov9(x[0], scale=lambda v: v * input_scale)
else:
tflite_common.set_input(interpreter, input)
interpreter.invoke()
objs = detect.get_objects(
interpreter, score_threshold=0.2, image_scale=(1, 1)
)
return objs
objs = await asyncio.get_event_loop().run_in_executor(self.executor, predict)