Bdtjtk

I managed to get the following code to take a list of images from a directory, apply object detection to them using the ImageAI Python module, and save the "processed" images to a new directory. This works great using the code below when I specify only a single additional thread to run it in.

When I attempt to increase the thread Range, it starts to throw this exception:

Cannot interpret feed_dict key as Tensor: Tensor
Tensor("Placeholder:0", shape=(7, 7, 3, 64), dtype=float32) is not an
element of this graph.

I wasn't able to find anything in the docs that says multithreading it isn't possible.

Some additional console output in case it's helpful (this is output when attempting 5 threads):

C:UsersMeDesktopProject>cd c:UsersMeDesktopProject && cmd /C "set "PYTHONIOENCODING=UTF-8" && set "PYTHONUNBUFFERED=1" && C:UsersMeAppDataLocalProgramsPythonPython36python.exe c:UsersMe.vscodeextensionsms-python.python-2018.12.1pythonFilesptvsd_launcher.py --default --client --host localhost

--port 61076 c:UsersMeDesktopProjectapp.py "

Using TensorFlow backend.

2019-01-02 21:52:37.108429: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1432] Found device 0 with properties:

name: GeForce GTX 1070 Ti major: 6 minor: 1 memoryClockRate(GHz): 1.683

pciBusID: 0000:03:00.0

totalMemory: 8.00GiB freeMemory: 6.61GiB

2019-01-02 21:52:37.113537: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0

2019-01-02 21:52:38.396715: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:

2019-01-02 21:52:38.399994: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988]      0

2019-01-02 21:52:38.402189: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0:   N

2019-01-02 21:52:38.404115: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6368

MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)

2019-01-02 21:52:38.419623: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0

2019-01-02 21:52:38.422946: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:

2019-01-02 21:52:38.425552: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988]      0

2019-01-02 21:52:38.427139: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0:   N

2019-01-02 21:52:38.428894: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6368

MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)

2019-01-02 21:52:38.436241: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0

2019-01-02 21:52:38.439749: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:

2019-01-02 21:52:38.442604: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988]      0

2019-01-02 21:52:38.444264: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0:   N

2019-01-02 21:52:38.445908: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6368

MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)

2019-01-02 21:52:38.460772: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0

2019-01-02 21:52:38.463329: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:

2019-01-02 21:52:38.466661: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988]      0

2019-01-02 21:52:38.468411: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0:   N

2019-01-02 21:52:38.470123: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6368

MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)

2019-01-02 21:52:38.476201: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0

2019-01-02 21:52:38.478613: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:

2019-01-02 21:52:38.484133: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988]      0

2019-01-02 21:52:38.493996: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0:   N

2019-01-02 21:52:38.496042: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6368

MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)

Backend TkAgg is interactive backend. Turning interactive mode on.

----Errors out on this line----

Working code for single thread:

import os

import cv2

import threading

import subprocess

import tensorflow as tf



from queue import Queue



from imageai.Detection import ObjectDetection



cwd = os.getcwd()

image_dir = os.path.join(cwd, 'TrainingImages')

processed_dir = os.path.join(cwd, 'ProcessedImages')



def process_image(image_filename):

    detector = ObjectDetection()

    detector.setModelTypeAsRetinaNet()

    detector.setModelPath(os.path.join(cwd, "resnet50_coco_best_v2.0.1.h5"))

    detector.loadModel(detection_speed="fast")



    if image_filename.endswith(".JPG"):

        try:

            detections = detector.detectObjectsFromImage(input_image=os.path.join(image_dir, image_filename), output_image_path=os.path.join(processed_dir, 'processed_' + image_filename))

            return detections

        except ValueError as e:

            print("Error processing: {}: {}".format(os.path.join(image_dir, image_filename), e))

            return False





def worker(q):

    ''' The worker thread pulls an item from the queue and processes it '''

    while True:

        image_filename = q.get()

        detections = process_image(image_filename)



        if detections:

            for obj in detections:

                print("{} : {}".format(obj["name"], obj["percentage_probability"]))

        else:

            print("No objects detected")



        q.task_done()





if __name__ == '__main__':

    # Get all files in image directory

    image_list = os.listdir(image_dir)



    # Create Queue for images

    q = Queue()



    for image_filename in image_list:

        q.put(image_filename)



    for i in range(1):

        t = threading.Thread(target=worker, args=(q, ))

        t.daemon = True

        t.start()



    q.join()

Project utilizes the following:

Python version 3.6.5

Package             Version

------------------- ---------

absl-py             0.6.1

astor               0.7.1

astroid             1.6.4

certifi             2018.4.16

chardet             3.0.4

colorama            0.3.9

cycler              0.10.0

gast                0.2.0

grpcio              1.17.1

h5py                2.9.0

idna                2.6

imageai             2.0.2

isort               4.3.4

Keras               2.2.4

Keras-Applications  1.0.6

Keras-Preprocessing 1.0.5

kiwisolver          1.0.1

lazy-object-proxy   1.3.1

Markdown            3.0.1

matplotlib          3.0.2

mccabe              0.6.1

numpy               1.15.4

opencv-python       3.4.5.20

Pillow              5.4.0

pip                 18.1

pipenv              2018.5.18

protobuf            3.6.1

pylint              1.9.1

pyparsing           2.3.0

python-dateutil     2.7.5

PyYAML              3.13

requests            2.18.4

scipy               1.2.0

setuptools          39.0.1

six                 1.11.0

tensorboard         1.12.1

tensorflow-gpu      1.12.0

termcolor           1.1.0

urllib3             1.22

virtualenv          16.0.0

virtualenv-clone    0.3.0

Werkzeug            0.14.1

wheel               0.32.3

wrapt               1.10.11

Question:

Can someone help me understand how to multithread the code above to allow processing multiple images asynchonously?

I tried to include as much info as possible.

Some additional info: It DOES appear to still run with multiple threads; however, every thread after the 1st throws that error above. It then continues on to process the first image, then essentially "hangs" (just doesn't do anything). I'm not familiar enough yet with the terminology yet to know what the Tensor "Graph" means or what to search for to troubleshoot further.

Here's the output with 5 threads:

2019-01-02 22:23:53.417469: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 6368 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070 Ti, pci bus id: 0000:03:00.0, compute capability: 6.1)

Error thrown: Cannot interpret feed_dict key as Tensor: Tensor Tensor("Placeholder:0", shape=(7, 7, 3, 64), dtype=float32) is not an element of this graph.

Error thrown: Cannot interpret feed_dict key as Tensor: Tensor Tensor("Placeholder:0", shape=(7, 7, 3, 64), dtype=float32) is not an element of this graph.

Error thrown: Cannot interpret feed_dict key as Tensor: Tensor Tensor("Placeholder:0", shape=(7, 7, 3, 64), dtype=float32) is not an element of this graph.

Error thrown: Cannot interpret feed_dict key as Tensor: Tensor Tensor("Placeholder:0", shape=(7, 7, 3, 64), dtype=float32) is not an element of this graph.

person : 78.54841947555542

truck : 56.28065466880798