luyishisi/anti-anti-spider

★ 7,295PythonAudience · researcherComplexity · 3/5Setup · hard

Mindmap

mindmap
  root((repo))
    What it does
      CAPTCHA recognition
      CNN training
      95 percent accuracy
    Architectures
      AlexNet
      LeNet
    Tech Used
      TensorFlow 1.9
      CPU or GPU
      JSON config
    Workflow
      Collect label images
      Run training script
      Test recognition
      Use pre-trained model

mindmap root((repo)) What it does CAPTCHA recognition CNN training 95 percent accuracy Architectures AlexNet LeNet Tech Used TensorFlow 1.9 CPU or GPU JSON config Workflow Collect label images Run training script Test recognition Use pre-trained model

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Learn how convolutional neural networks work in practice by training one on a concrete image-recognition task like CAPTCHA solving.

USE CASE 2

Compare AlexNet and LeNet architectures on a small image dataset and observe the accuracy and speed trade-offs firsthand.

USE CASE 3

Download the pre-trained model to run CAPTCHA recognition in a research or learning context without training from scratch.

Tech stack

PythonTensorFlowCNNAlexNetLeNetCUDA

Getting it running

Difficulty · hard Time to first run · 1day+

Requires TensorFlow 1.9 (old version), labeled CAPTCHA images to train on, and an NVIDIA GPU for reasonable training speed.

No license information was mentioned in the explanation.

In plain English

This repository contains a Python library for automatically reading and solving CAPTCHA images using a type of neural network called a convolutional neural network (CNN). CAPTCHAs are the distorted text or image puzzles that websites use to check whether a visitor is a human rather than an automated program. This project was built as a learning resource around the specific technical challenge of training a model to recognize those images. The README is written in Chinese. It describes using two CNN model architectures called AlexNet and LeNet, and reports around 95.5% accuracy on CAPTCHA recognition. The training process runs on TensorFlow 1.9.0, either on a standard CPU or with an NVIDIA GPU for faster training. AlexNet requires images resized to 227 by 227 pixels, and the repository includes a preprocessing script to handle that resizing. The workflow is: collect and label CAPTCHA images, split them into training and validation sets, place them in the sample directory, run the training script, and then use the recognition script to test how well the trained model performs. Configuration is handled through a JSON file. The author notes the project is now paused due to other commitments, and emphasizes it is intended for learning about image recognition and CNNs only, not for malicious use. The project folder also includes older content in an Anti-Anti-Spider subdirectory covering related web-scraping techniques. The project description and topics are in Chinese, and the README does not include an English version. Pre-trained model files are available for download from a link in the README if training from scratch is not needed.

Copy-paste prompts

Prompt 1

I cloned luyishisi/anti-anti-spider and want to train the AlexNet model on my own CAPTCHA dataset. Walk me through labeling images, setting up the sample directory, and running the training script.

Prompt 2

The anti-anti-spider project uses TensorFlow 1.9. Help me set up a compatible Python virtual environment with the right TensorFlow version on a machine without a GPU.

Prompt 3

I want to use the pre-trained CAPTCHA model from luyishisi/anti-anti-spider to recognize images in a Python script. Show me how to load the model and run inference on a single image.

Prompt 4

Explain the difference between the AlexNet and LeNet architectures used in this CAPTCHA project, why does AlexNet need 227x227 images and how does that affect accuracy versus speed?

Open on GitHub → Explain another repo

← luyishisi on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.