Using Image url in Keras image classification

 
How to load an image in Keras for machine learning from a url ?

For loading image in Keras we used to write code like this

img = keras.utils.load_img(path="OIP.jpg",
color_mode = "grayscale",target_size=(28,28,1))

How about using a image URL instead of the above code?

Need to perform couple of things.

  1. Get response from the image url.
  2. Get the image as byte using ByteIO
  3. Covert the image as Grayscale for prediction and apply appropriate resize.
  4. Finally covert the image into a NumPy array.
Let's wrap all these in a function.

from tensorflow import keras
from io import BytesIO
from PIL import Image
import requests

def loadImage(url):
response = requests.get(url)
img_bytes = BytesIO(response.content)
img = Image.open(img_bytes)
img = img.convert('L')
# img = img.convert('1')
 # img = img.convert('RGB')
 img = img.resize((28, 28), Image.NEAREST)
img = keras.utils.img_to_array(img)

return img


Before using the image for prediction, we need to reshape the image.

url = 'https://edwin-de-jong.github.io/blog/mnist-sequence-data/fig/5.png'
img = loadImage(url)
test_img = img.reshape((1, 784))
img_class = model.predict(test_img)

Comments

Post a Comment

Popular posts from this blog

Zustland : a minimal data store to consider for the Next React Project

 I have been depending on Context for passing state through components in React and eventually things complicated when data is become complex. Move on to beautiful Redux but it is much complex to digest. For learning Redux I have spent lots of time and Redux Toolkit, easy to use Redux, sav ed me a lot. Even though RTK is ok, it is not a simple library. I am fan of Svelte store and Vuex , Pinia store. Is there any such library for React ? I recently found Zustand a minimal store for React. Love it. 😍 Create a store Binding Updating the state Create a store Create a store using the create an d export the custom hook. <pre class='language-javascript'> <code> import create from 'zustand' const useBearStore = create((set) => ({ bears: 0, increasePopulation: () => set((state) => ({ bears: state.bears + 1 })), removeAllBears: () => set({ bears: 0 }), })) </code></pre> Binding In your component use the ho...
Read more

How to run HugginFace models in Python

  Hugging Face Hugging face is a machine learning community where you can share pre trained models and explore other trained models, which are suited for many use cases. Many of the models already trained well and ready to use. Without delay get started! Python setup Open your Jupiter notebook or for easy startup without python installation on local machine, use Colab . In this example we are using mistralai/Mistral-7B-v0.1 model. Let's install the transformers module which is the primary dependency for HuggingFace.   conda install transformers Following script will work like a charm . # mistrel model test from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "mistralai/Mistral-7B-v0.1" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id) text = "Hello my name is" inputs = tokenizer(text, return_tensors="pt") outputs = model.generate(**inputs, max_new_tokens=20) print(tokenizer.d...
Read more