EDCC: An efficient and accurate algorithm for palmprint-recognition

EDCC(Enhanced and Discriminative Competitive Code), which is used for palmprint-recognition.

Use the EDCC algorithm with default config to validate on several published palmprint databases(multispectral, tongji), the first N(N = 2, 4, 6, 8) palmprint images of each palm are employed as training samples and the remaining palmprint images form the test sample set. Each sample in the test sample set is compared with all samples of each class in the training set to calculate the matching score. The class that produces the highest matching score is treated as the class of the test sample.

Advantages of EDCC algorithm:

1. Less training samples.
2. Faster recognition speed.
3. Higher recognition accuracy.

More details about EDCC

Installation

Install library

There are some requirements if you want to install EDCC library:

• OS *nix like.
• OpenCV 3.4 or OpenCV 4.5 installed.

Steps:

1. git clone https://github.com/Leosocy/EDCC-Palmprint-Recognition.git
2. cd EDCC-Palmprint-Recognition && mkdir -p build && cd build
3. cmake .. && sudo make install

Install Python Package

Please make sure that the edcc library has been successfully installed by following the steps above.

Python3.x required.

Steps:

1. cd pypackage
2. python setup.py install

QuickStart

The project provides a Docker container runtime environment with edcc library and python package installed.

You can quick start accord to the following commands:

# bootstrap a docker container with edcc library installed
./manage.sh env

# run c example
cd /app/examples/c_example && mkdir -p build && cd build && cmake .. && make && ./run_c_sample

# run cpp example
cd /app/examples/cpp_example && mkdir -p build && cd build && cmake .. && make && ./run_cpp_sample

# run python example
cd /app/examples/py_example && python example.py

Usage

Make sure you have installed library and Python package before using edcc.

And you can see more usage details under examples directory about usage.

C/C++

In your CMakeLists.txt, add these lines:

find_package(edcc REQUIRED)
include_directories(${EDCC_INCLUDE_DIR})
...
add_dependencies(${YOUR_PROJECT} ${EDCC_LIBRARIES})
target_link_libraries(${YOUR_PROJECT} ${EDCC_LIBRARIES})

Then you can use it in your source code(C or C++) like this:

C

#include <edcc/c_api.h>

#define ASSERT_STATUS_OK(s) \
  do {                      \
    if (s[0] != '\0') {     \
      perror(s + 1);        \
      return -1;            \
    }                       \
  } while (0)

int main() {
  // create a new encoder.
  char status[128];
  int encoder_id = new_encoder_with_config(29, 5, 5, 10, status);
  ASSERT_STATUS_OK(status);
  // encode palmprints to code buffer.
  unsigned long buffer_size = get_size_of_code_buffer_required(encoder_id);
  char* code_buffer_one = (char*)malloc(buffer_size);
  char* code_buffer_another = (char*)malloc(buffer_size);
  encode_palmprint_using_file(encoder_id, one_image_file_path, code_buffer_one, buffer_size, status);
  ASSERT_STATUS_OK(status);
  encode_palmprint_using_file(encoder_id, another_image_file_path, code_buffer_another, buffer_size, status);
  ASSERT_STATUS_OK(status);
  // calculate the similarity score of two codes.
  double score = calculate_codes_similarity(code_buffer_one, code_buffer_another, status);
  ASSERT_STATUS_OK(status);
  return 0;
}

C++

#include <edcc/facade.h>
#include <edcc/status.h>

#define ASSERT_STATUS_OK(s) \
  do {                      \
    if (!s.IsOk()) {        \
      perror(s.msg());      \
      return -1;            \
    }                       \
  } while (0)

using edcc::EdccFacade;
using edcc::Status;

int main() {
  Status s;
  // create a new encoder.
  auto inst = EdccFacade::Instance();
  auto encoder_id = inst->NewEncoderWithConfig(29, 5, 5, 10, &s);
  ASSERT_STATUS_OK(s);
  // encode palmprints to code buffer.
  size_t buffer_size = inst->GetSizeOfCodeBufferRequired(encoder_id);
  char* code_buffer_one = new char[buffer_size];
  char* code_buffer_another = new char[buffer_size];
  inst->EncodePalmprint(encoder_id, one_image_file_path, code_buffer_one, buffer_size, &s);
  ASSERT_STATUS_OK(s);
  inst->EncodePalmprint(encoder_id, another_image_file_path, code_buffer_another, buffer_size, &s);
  ASSERT_STATUS_OK(s);
  // calculate the similarity score of two codes.
  double score = inst->CalcCodeSimilarity(code_buffer_one, code_buffer_another, &s);
  ASSERT_STATUS_OK(s);
  return 0;
}

Python

import edcc

config  = edcc.EncoderConfig(29, 5, 5 ,10)
encoder = edcc.create_encoder(config)
one_palmprint_code = encoder.encode_using_filename("./palmprint_one.bmp")
another_palmprint_code = encoder.encode_using_filename("./palmprint_another.bmp")
similarity_score = one_palmprint_code.compare_to(another_palmprint_code)

Contributing

Please see CONTRIBUTING.md