I did create a machine learning model using Pytorch which i want to use as a webservice using Flask. The problem is that i don't understand how i can pass a json-String to the url. Below is my code that I wrote to do some tryouts with my model and Flask:
from modelLoader import Model
from imageLoader import Img
import os
from flask import Flask, jsonify, request
app = Flask(__name__)
classes = ["dummy-image", "product-image"]
model_path = os.path.join("data", "models", "model1709", "model1709")
image_path = os.path.join("data", "images", "dummy_images")
m1 = Model(model_path, classes, "cpu")
#app.route('/predict', methods=['POST', 'GET'])
def predict():
# case for handle json
input_data = request.get_json()['url']
if isinstance(input_data, list):
for elem in input_data:
img_elem = Img(url=elem)
res = img_elem.get_prediction(m1)
return jsonify({"type": "bulk_upload"})
img_inpdata = Img(url=input_data)
res, info = img_inpdata.get_prediction(m1)
return jsonify({input_data: res, "info": str(info)})
if __name__ == '__main__':
app.run(debug=True)
This would be a request that I want to make using this code:
POST http://192.168.178.13:5000/predict HTTP/1.1
Content-Type: application/json
Accept: application/json
{
"url" : "https://socialistmodernism.com/wp-content/uploads/2017/07/placeholder-image.png"
}
How exactly can I get the prediction for the image inside the json-string, by passing this json-string to the application?
Here the two classes model and imageLoader for completeness:
from torch import argmax, device, load, nn
class Model:
def __init__(self, path, class_list=None, dvc=None):
if class_list is None:
class_list = [0, 1]
if dvc is None:
dvc = 'cpu'
self.class_list = class_list
self.model = load(path, map_location=device(dvc))
num_ftrs = self.model.fc.in_features
self.model.fc = nn.Linear(num_ftrs, len(class_list))
self.model.eval()
import torchvision.transforms as transforms
import io
from PIL import Image
from torch import argmax, device, load, nn
import requests
class Img:
def __init__(self, url=None, image=None, image_bytes=None):
if url:
img = Image.open(requests.get(url, stream=True).raw)
img_byte_arr = io.BytesIO()
img.save(img_byte_arr, format=img.format)
self.image_bytes = img_byte_arr.getvalue()
elif image:
f = image.read()
self.image_bytes = bytearray(f)
elif image_bytes:
self.image_bytes = image_bytes
def transform_image(self):
data_transforms = transforms.Compose([transforms.Resize((224, 224)),
transforms.CenterCrop(
224), transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224,
0.225])])
image = Image.open(io.BytesIO(self.image_bytes)).convert('RGB')
return data_transforms(image).unsqueeze(0)
def get_prediction(self, model):
tensor = self.transform_image()
output = (model.model(tensor))
sm = nn.Softmax(output)
best = output.argmax().item()
return model.class_list[best], sm
You cannot do a POST request directly from the browser URL box. There are many applications to test your API and my favorite one is postman. You can also use the curl command tool.
If you want to use the browser URL box only, then consider using the GET request. The format of GET request is <URL>?parameter1=value1¶meter2=value2. You can access the value of the parameter in flask using the request module. For example, if your service is at http://192.168.178.13:5000/predict. You can send it as http://192.168.178.13:5000/predict?url=your-url. And you can fetch it in flask as
from flask import request
my_url = request.args.get("url")
Related
I am trying to make a website that can make predictions on images using tensorflow, flask, and python.
This is my code:
from flask import Flask, render_template
import os
import numpy as np
import pandas as pd
app = Flask(__name__)
#app.route('/')
def index():
return render_template('index.html')
import tensorflow as tf
import tensorflow_hub as hub
model = tf.keras.models.load_model(MODEL_PATH)
IMG_SIZE = 224
BATCH_SIZE = 32
custom_path = "http://t1.gstatic.com/licensed-image?q=tbn:ANd9GcQd6lM4HtInRF3cxw6h3MgUZIIiJCdMgFvXKrhaJrbw61tN3aYpMIVBi0dx0KPv1sdCrLk0sBhPeNVt8m0"
custom_data = create_data_batches(custom_path, test_data=True)
custom_preds = model.predict(custom_data)
# Get custom image prediction labels
custom_pred_labels = [get_pred_label(custom_preds[i]) for i in range(len(custom_preds))]
print(custom_pred_labels)
#app.route('/my-link/')
def my_link():
return f"The predictions are: {custom_pred_labels}"
if __name__ == '__main__':
app.run(host="localhost", port=3000, debug=True)
The process_image function:
def process_image(image_path, img_size=IMG_SIZE):
"""
Takes an image file path and turns the image into a Tensor.
"""
image = tf.io.read_file(image_path)
image = tf.image.decode_jpeg(image, channels=3)
image = tf.image.convert_image_dtype(image, tf.float32)
image = tf.image.resize(image, size=[img_size, img_size])
return image
The needed part of the create_data_batches function:
def create_data_batches(X, y=None, batch_size=BATCH_SIZE, valid_data=False, test_data=False):
"""
Creates batches out of data out of image (X) and label (y) pairs.
Shuffles the data if it's training data but doesn't shuffle if it's validation data.
Also accepts test data as input (no labels)
"""
if test_data:
print("Creating test data batches...")
data = tf.data.Dataset.from_tensor_slices((tf.constant(X))) # only filepaths (no labels)
data_batch = data.map(process_image).batch(BATCH_SIZE)
return data_batch
The get_image_label function:
def get_image_label(image_path, label):
"""
Takes an image file path name and the associated label, processes the image and returns a tuple of (image, label).
"""
image = process_image(image_path)
return image, label
The get_pred_label function:
def get_pred_label(prediction_probabilites):
"""
Turns an array of prediction probabilities into a label.
"""
return unique_breeds[np.argmax(prediction_probabilites)]
Now when I run this, I get the following error:
ValueError: Unbatching a tensor is only supported for rank >= 1
I tried turning it into a list as one of the solutions I found said:
custom_path = ["http://t1.gstatic.com/licensed-image?q=tbn:ANd9GcQd6lM4HtInRF3cxw6h3MgUZIIiJCdMgFvXKrhaJrbw61tN3aYpMIVBi0dx0KPv1sdCrLk0sBhPeNVt8m0"]
But when I run that, I get this error:
UNIMPLEMENTED: File system scheme 'http' not implemented (file: 'http://t1.gstatic.com/licensed-image?q=tbn:ANd9GcQd6lM4HtInRF3cxw6h3MgUZIIiJCdMgFvXKrhaJrbw61tN3aYpMIVBi0dx0KPv1sdCrLk0sBhPeNVt8m0')
Any help would be appreciated.
I have an Image object that is required to be passed to the client side and display on the page. Below is what I tried but to not avail.
The model_predict function is receiving the image from client for classification. After the classification has been done then the model will generate an image to be passed back to the client in function upload().
def model_predict(img_path, model):
img = image.load_img(img_path, target_size=(224, 224))
...
leaf_result = cv2.bitwise_and(img, img, mask=mask)
im = Image.fromarray(leaf_result)
# decoded_class = decode(img)
im.save(f"./output.png")
preds = model.predict(x)
return preds,im
#app.route('/predict', methods=['GET', 'POST'])
def upload():
if request.method == 'POST':
# Get the file from post request
f = request.files['file']
# Save the file to ./uploads
basepath = os.path.dirname(__file__)
file_path = os.path.join(
basepath, 'uploads', secure_filename(f.filename))
f.save(file_path)
prediction, im = model_predict(file_path, model)
....
result_list = list()
result_list.append(prediction1)
result_list.append(prediction2)
return jsonify(result_list, base64.decodebytes(im))
return None
Base64 Encoding is what you need: send base64 data to client, then your client will decode it. Flask to return image
For e.g:
import io
import base64
# import flask
from PIL import Image
def get_encoded_img(image_path):
img = Image.open(image_path, mode='r')
img_byte_arr = io.BytesIO()
img.save(img_byte_arr, format='PNG')
my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
return my_encoded_img
...
# your api code
...
img_path = 'assets/test.png'
img = get_encoded_img(img_path)
# prepare the response: data
response_data = {"key1": value1, "key2": value2, "image": img}
# return flask.jsonify(response_data )
When your client receives base64 data, it can display that image by using naive JS. How to display Base64 images in HTML
For e.g:
<img id="image"/>
<script>
var ticket = {
"validationDataValidDate": "2019-05-30",
"validationImage": "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"
};
document.querySelectorAll('[id="image"]')[0].src = 'data:image/png;base64, '+ticket["validationImage"];
</script>
trying to use tensorflow on ipage server with operating system centOs 7
I don't know if there is a GPU or not , but I get this error message
kernel driver does not appear to be running on this host /proc/driver/nvidia/version does not exist
I tried this code
from tensorflow import config
config.set_soft_device_placement = True
and also this code trying to prevent tensorflow from using GPU
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
but I kept receiving the same error message
Note :
I am trying to run face recognition script on server to be used by php through terminal using exec() function by passing the image as string base64 and then decoding it and then adding new face or fetching faces and comparing and then will be saved in the database
and this is my script :
import face_recognition as fr
from tensorflow.keras.models import model_from_json
import numpy as np
from PIL import Image
import base64
import io
import json
import cv2
from tensorflow import config
config.set_soft_device_placement = True
def check_if_spoof(image_string_base64):
# decoding the image
msg = base64.b64decode(image_string_base64)
buf = io.BytesIO(msg)
img = Image.open(buf).convert("RGB")
opencv_image = np.array(img)
test_image = opencv_image[:,:,::-1].copy()
json_file = open('antispoofing_models/antispoofing_model.json','r')
loaded_model_json = json_file.read()
json_file.close()
model = model_from_json(loaded_model_json)
# load antispoofing model weights
model.load_weights('antispoofing_models/antispoofing_model.h5')
resized_face = cv2.resize(test_image,(160,160))
resized_face = resized_face.astype("float") / 255.0
# resized_face = img_to_array(resized_face)
resized_face = np.expand_dims(resized_face, axis=0)
# pass the face ROI through the trained liveness detector
# model to determine if the face is "real" or "fake"
preds = model(resized_face)[0]
if preds> 0.5:
return True
else:
return False
def decode_img(image_string_base64):
msg = base64.b64decode(image_string_base64)
buf = io.BytesIO(msg)
img = Image.open(buf)
return img
def resize_image(image_string_base64, height=500):
image = decode_img(image_string_base64)
height_percent = (height / float(image.size[1]))
width_size = int((float(image.size[0]) * float(height_percent)))
image = image.resize((width_size, height), Image.NEAREST)
return np.array(image)
def add_person(image_string_base64):
if check_if_spoof(image_string_base64):
return {'error':'image is fake'}
img = np.array(decode_img(image_string_base64))
img_encodings = fr.face_encodings(img, num_jitters=4)
return np.array(img_encodings)
def get_person(image_string_base64, data, tolerance=0.35):
if check_if_spoof(image_string_base64):
return {'error':'image is fake'}
data = json.loads(data)
resized_img = resize_image(image_string_base64)
img_encoding = fr.face_encodings(resized_img)
############## testing purpose only #########
# return fr.compare_faces(data,img_encoding)
#############################################
# type of recieved data is : { id ->number : encoding->list}
for user in data:
for id_ , encoding in user:
if True in fr.compare_faces(encoding, img_encoding):
return id_
return {'error':"user not existed"}
I'm successfully trained my own dataset using Keras yolov3 Github project link
and I've got good predictions:
I would like to deploy this model on the web using flask to make it work with a stream or with IP cameras.
I saw many tutorials explains how to do that but, in reality, I did not find what I am looking for.
How can I get started?
You can use flask-restful to design a simple rest API.
You can use opencv VideoCapture to grab the video stream and get frames.
import numpy as np
import cv2
# Open a sample video available in sample-videos
vcap = cv2.VideoCapture('URL')
The client will take an image/ frame, encode it using base64, add other details like height, width, and make a request.
import numpy as np
import base64
import zlib
import requests
import time
t1 = time.time()
for _ in range(1000): # 1000 continuous request
frame = np.random.randint(0,256, (416,416,3), dtype=np.uint8) # dummy rgb image
# replace frame with your image
# compress
data = frame # zlib.compress(frame)
data = base64.b64encode(data)
data_send = data
#data2 = base64.b64decode(data)
#data2 = zlib.decompress(data2)
#fdata = np.frombuffer(data2, dtype=np.uint8)
r = requests.post("http://127.0.0.1:5000/predict", json={'imgb64' : data_send.decode(), 'w': 416, 'h': 416})
# make a post request
# print the response here
t2 = time.time()
print(t2-t1)
Your server will load the darknet model, and when it receives a post request it will simply return the model output.
from flask import Flask, request
from flask_restful import Resource, Api, reqparse
import json
import numpy as np
import base64
# compression
import zlib
# load keras model
# load_model('model.h5')
app = Flask(__name__)
api = Api(app)
parser = reqparse.RequestParser()
parser.add_argument('imgb64', location='json', help = 'type error')
parser.add_argument('w', type = int, location='json', help = 'type error')
parser.add_argument('h', type = int, location='json', help = 'type error')
class Predict(Resource):
def post(self):
request.get_json(force=True)
data = parser.parse_args()
if data['imgb64'] == "":
return {
'data':'',
'message':'No file found',
'status':'error'
}
img = data['imgb64']
w = data['w']
h = data['h']
data2 = img.encode()
data2 = base64.b64decode(data2)
#data2 = zlib.decompress(data2)
fdata = np.frombuffer(data2, dtype=np.uint8).reshape(w, h, -1)
# do model inference here
if img:
return json.dumps({
'mean': np.mean(fdata),
'channel': fdata.shape[-1],
'message':'darknet processed',
'status':'success'
})
return {
'data':'',
'message':'Something when wrong',
'status':'error'
}
api.add_resource(Predict,'/predict')
if __name__ == '__main__':
app.run(debug=True, host = '0.0.0.0', port = 5000, threaded=True)
In the # do model inference here part, just use your detect/predict function.
If you want to use native darknet, https://github.com/zabir-nabil/tf-model-server4-yolov3
If you want to use gRPC instead of REST, https://github.com/zabir-nabil/simple-gRPC
I've deployed a custom Pytorch model to the Google AI platform for prediction, but when I try to make a prediction request with image data using gcloud tools I get the following error in response:
{
"error": "Prediction failed: unknown error."
}
I've tried to encode my image data in b64 format or to place it into a multidimensional python array, by doing the following:
pil_im = Image.open('Pic512.png')
pil_im = pil_im.resize((224,224)).convert('RGB')
im_arr = np.asarray(pil_im)
py_arr = im_arr.tolist()
json_instance_1 = {'instances': py_arr}
with open('json_instance_1.json', 'w') as f:
json.dump(json_instance_1, f)
I converted it into b64 like so, after adjusting my Predictor code accordingly:
with open('Pic512.png', 'rb') as f:
byte_im = f.read()
json_instance = {'instances': {'b64': base64.b64encode(byte_im).decode()}}
with open('json_instance.json', 'w') as f:
json.dump(json_instance, f)
I've tried converting with different file formats and similar methods, but all of them give me the same error.
My predictor module:
from facenet_pytorch import MTCNN, InceptionResnetV1, extract_face
import torch
from torchvision import transforms
from torch.nn import functional as F
from PIL import Image
# from sklearn.externals import joblib
import numpy as np
import os
import io
import base64
class MyPredictor(object):
"""An example Predictor for an AI Platform custom prediction routine."""
def __init__(self, model, preprocessor, device):
"""Stores artifacts for prediction. Only initialized via `from_path`.
"""
self._resnet = model
self._mtcnn_mult = preprocessor
self._device = device
self.get_std_tensor = transforms.Compose([
np.float32,
np.uint8,
transforms.ToTensor(),
])
self.tensor2pil = transforms.ToPILImage(mode='RGB')
self.trans_resnet = transforms.Compose([
transforms.Resize((100, 100)),
np.float32,
transforms.ToTensor()
])
def predict(self, instances, **kwargs):
pil_transform = transforms.Resize((512, 512))
imarr = np.uint8(np.array(instances))
# img_bytes_string = io.BytesIO(base64.b64decode(instances))
pil_im = Image.fromarray(imarr)
# pil_im = Image.open(img_bytes_string)
image = pil_im.convert('RGB')
pil_im_512 = pil_transform(image)
boxes, _ = self._mtcnn_mult.detect(pil_im_512)
box = boxes[0]
face_tensor = extract_face(pil_im_512, box, margin=40)
std_tensor = self.get_std_tensor(face_tensor.permute(1, 2, 0))
cropped_pil_im = self.tensor2pil(std_tensor)
face_tensor = self.trans_resnet(cropped_pil_im)
face_tensor4d = face_tensor.unsqueeze(0)
face_tensor4d = face_tensor4d.to(self._device)
self._resnet.eval()
prediction = self._resnet(face_tensor4d)
preds = F.softmax(prediction, dim=1).detach().numpy().reshape(-1)
print('probability of (class1, class2) = ({:.4f}, {:.4f})'.format(preds[0], preds[1]))
return {'probs':preds.tolist()}
#classmethod
def from_path(cls, model_dir):
device_path = os.path.join(model_dir, 'device_cpu.pt')
device = torch.load(device_path)
model_path = os.path.join(model_dir, 'FullResNetRefinedExtra_no_norm_100x100_8634.pt')
classifier = torch.load(model_path, map_location=device)
mtcnn_path = os.path.join(model_dir, 'mtcnn_mult.pt')
mtcnn_mult = torch.load(mtcnn_path)
return cls(classifier, mtcnn_mult, device)
When I test the class locally everything works, so I assume it's a problem related the serialisation and deserialisation on the side of Google Platform. How can I resolve this issue?