Skip to content

DeepAPI v0.4.3

Latest
Compare
Choose a tag to compare
@wuhanstudio wuhanstudio released this 29 May 15:48
· 4 commits to master since this release

Deep Learning as a Cloud API Service.

Quick Start:

# Using docker
docker run -p 8080:8080 wuhanstudio/deepapi

# Using python3
pip install deepapi
python -m deepapi

Serving on port 8080...

The cloud service is available at http://localhost:8080

Changelog

Supported models for research on black-box adversarial attacks,:

  • Pretrained VGG16 model on Cifar10 dataset
  • Pretrained VGG16 model on ImageNet dataset
  • Pretrained Resnet50 model on ImageNet dataset
  • Pretrained Inceptionv3 model on ImageNet dataset

The Web UI:

  • Upload and recognize an image
  • Automatic python client code generation
  • Automatic curl client code generation

The Python API:

import numpy as np
from PIL import Image

from deepapi.api import DeepAPI_VGG16_Cifar10

# Load the image
x = Image.open("dog.jpg")
x = np.array(x)

# Initialize the model
model  = DeepAPI_VGG16_Cifar10('http://localhost:8080', concurrency=8)

# Predict
y = model.predict(np.array([x]))[0]

# Print the result
model.print(y)