Custom implementation

The implementation of this fork for a custom dataset can be found in this Google Colab notebook.

Objects counting by estimating a density map with convolutional neural networks

The PyTorch implementation of https://github.com/WeidiXie/cell_counting_v2.

More details about the method can be found in our blog post.

Data preparation

Three datasets are considered:

• cell: Fluorescent cells dataset
• mall: Mall dataset
• ucsd: UCSD pedestrians dataset

One can get them using get_data.py script:

Usage: get_data.py [OPTIONS]

  Get chosen dataset and generate HDF5 files with training and validation samples.

Options:
  --dataset [cell|mall|ucsd]  [required]
  --help                      Show this message and exit.

The script download original data and preprocess it to create HDF5 files with images and labels (corresponding density maps).

Training

Two network architectures are available so far:

• U-Net

  Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. "U-net: Convolutional networks for biomedical image segmentation." In International Conference on Medical image computing and computer-assisted intervention, pp. 234-241. Springer, Cham, 2015.

• FCRN-A

  Weidi, Xie, J. Alison Noble, and Andrew Zisserman. "Microscopy cell counting with fully convolutional regression networks." In 1st Deep Learning Workshop, Medical Image Computing and Computer-Assisted Intervention (MICCAI). 2015.

One can use train.py script to train chosen network on selected dataset:

Usage: train.py [OPTIONS]

  Train chosen model on selected dataset.

Options:
  -d, --dataset_name [cell|mall|ucsd]
                                  Dataset to train model on (expect proper
                                  HDF5 files).  [required]
  -n, --network_architecture [UNet|FCRN_A]
                                  Model to train.  [required]
  -lr, --learning_rate FLOAT      Initial learning rate (lr_scheduler is
                                  applied).
  -e, --epochs INTEGER            Number of training epochs.
  --batch_size INTEGER            Batch size for both training and validation
                                  dataloaders.
  -hf, --horizontal_flip FLOAT    The probability of horizontal flip for
                                  training dataset.
  -vf, --vertical_flip FLOAT      The probability of horizontal flip for
                                  validation dataset.
  --unet_filters INTEGER          Number of filters for U-Net convolutional
                                  layers.
  --convolutions INTEGER          Number of layers in a convolutional block.
  --plot                          Generate a live plot.
  --help                          Show this message and exit.

Inference

The infer.py script is provided to run a model on a given input image.

Usage: infer.py [OPTIONS]

  Run inference for a single image.

Options:
  -i, --image_path FILENAME       A path to an input image.  [required]
  -n, --network_architecture [UNet|FCRN_A]
                                  Model architecture.  [required]
  -c, --checkpoint FILENAME       A path to a checkpoint with weights.
                                  [required]

  --unet_filters INTEGER          Number of filters for U-Net convolutional
                                  layers.

  --convolutions INTEGER          Number of layers in a convolutional block.
  --one_channel                   Turn this on for one channel images
                                  (required for ucsd).

  --pad                           Turn on padding for input image (required
                                  for ucsd).

  --visualize                     Visualize predicted density map.
  --help                          Show this message and exit.

Examples

$ python infer.py -n UNet -c cell_UNet.pth -i ../data/cells/001cell.png --visualize                                    
The number of objects found: 134.857265625

python infer.py -n UNet -c mall_UNet.pth -i ../data/mall_dataset/frames/seq_000001.jpg --visualize                                    
The number of objects found: 30.82901123046875

python infer.py -n UNet -c ucsd_UNet.pth -i ../data/ucsdpeds/vidf/vidf1_33_000.y/vidf1_33_000_f001.png --pad --one_channel --visualize
The number of objects found: 18.083719482421873

Requirements

To install required python packages run: pip3 install -r requirements.txt.

---

If you are interested in ML/DL methods, check out our blog for more articles.