animl 3.3.0

A Collection of ML Tools for Species Detection and Classification in Camera Trap Images and Videos.

animl-py 3.3.0

AniML comprises a variety of machine learning tools for analyzing ecological data. This Python package includes a set of functions to classify subjects within camera trap field data and can handle both images and videos. This package is also available in R: animl

Table of Contents

1. Installation
2. Usage
3. Models

Installation Instructions

It is recommended that you set up a conda environment for using animl. See Dependencies below for more detail. You will have to activate the conda environment first each time you want to run AniML from a new terminal.

From GitHub

git clone https://github.com/conservationtechlab/animl-py.git
cd animl-py
pip install -e .

From PyPi

pip install animl

Dependencies

We recommend running AniML on GPU-enabled hardware. **If using an NVIDIA GPU, ensure driviers, cuda-toolkit and cudnn are installed.

Python >= 3.12

PyTorch
Animl currently depends on torch >= 2.6.0. To enable GPU, install the CUDA-enabled version

Python Package Dependencies

• dill>=0.4.0
• numpy>=2.0.2
• onnxruntime-gpu>=1.19.2
• pandas>=2.2.2, <3.0.0
• pillow>=11.0.0
• pyexiftool>=0.5.6
• opencv-python>=4.12.0.88
• scikit-learn>=1.5.2
• timm>=1.0.9
• torch>=2.6.0
• torchvision>=0.21.0
• tqdm>=4.66.5
• ultralytics>=8.3.95
• wget>=3.2

Verify Install

We recommend you download the examples folder within this repository. Download and unarchive the zip folder. Then with the conda environment active:

python -m animl /path/to/example/folder

This should create an Animl-Directory subfolder within the example folder. Add -s to sort the images into species folders. Add -v to add copies of the images with the bounding boxes plotted.

Or, if using your own data/models, animl can be given the paths to those files: Download and unarchive the zip folder. Then with the conda environment active:

python -m animl /example/folder --detector /path/to/megadetector --classifier /path/to/classifier --classlist /path/to/classlist.txt

You can use animl in this fashion on any image directory.

Finally you can use the animl.yml config file to specify parameters:

python -m animl /path/to/animl.yml

Usage

See the Reference Manual for detailed information on functions and usage.

Inference

The functionality of animl can be parcelated into its individual functions to suit your data and scripting needs. The sandbox.ipynb notebook has all of these steps available for further exploration.

1. It is recommended that you use the animl working directory for storing intermediate steps.

   import animl
   workingdir = animl.WorkingDirectory('/path/to/save/data')
   

2. Build the file manifest of your given directory. This will find both images and videos.

   files = animl.build_file_manifest('/path/to/images', out_file=workingdir.filemanifest, exif=True, data_timezone='America/Los_Angeles')
   

3. If there are videos, extract individual frames for processing. Select either the number of frames or fps using the argumments. The other option can be set to None or removed.

   allframes = animl.extract_frames(files, frames=3, out_file=workingdir.imageframes, parallel=True)
   

4. Pass all images into MegaDetector. We recommend MDv5a. The function parse_MD will convert the json to a pandas DataFrame and merge detections with the original file manifest, if provided.

   detector = animl.load_detector('/path/to/mdmodel.pt', model_type="mdv5", device='cuda:0')
   mdresults = animl.detect(detector,
                           allframes,
                           resize_width=animl.MEGADETECTORv5_SIZE,
                           resize_height=animl.MEGADETECTORv5_SIZE, 
                           letterbox=True,
                           device='cuda:0',
                           batch_size = 4,
                           checkpoint_path=working_dir.mdraw,
                           checkpoint_frequency=1000)
   detections = animl.parse_detections(mdresults, manifest=allframes, out_file=workingdir.detections)
   

5. For speed and efficiency, extract the empty/human/vehicle detections before classification.

   animals = animl.get_animals(detections)
   empty = animl.get_empty(detections)
   

6. Classify using the appropriate species model. Merge the output with the rest of the detections if desired.

   classifier, class_list = animl.load_classifier('/path/to/model', '/path/to/classlist.txt', device='cuda:0')
   predictions_output = animl.classify(classifier,
                                    animals,
                                    resize_width=480,
                                    resize_height=480,
                                    batch_size=4,
                                    out_file=working_dir.predictions)
   

7. Apply labels from class list with or without utilizing timestamp-based sequences.

   manifest = animl.single_classification(animals, empty, predictions_output, class_list['class'])
   

   or, after defining a station column,

   manifest = animl.sequence_classification(animals,
                                            empty, 
                                            predictions_output,
                                            class_list['class'],
                                            station_col='station',
                                            empty_class="",
                                            sort_columns=None,
                                            file_col="filepath",
                                            maxdiff=60)
   

8. (OPTIONAL) Save the Pandas DataFrame's required columns to csv and then use it to create json for TimeLapse compatibility

   csv_loc = animl.export_timelapse(manifest, 'animl_export/timelapse/', only_animal = True)
   animl.export_megadetector(manifest, out_file ="final_result.json", detector = 'MegaDetector v5a')
   

9. (OPTIONAL) Create symlinks within a given directory for file browser access.

   manifest = animl.export_folders(manifest, out_dir=working_dir.linkdir, out_file=working_dir.results)
   

---

Training

Training workflows are still under development. Please submit Issues as you come upon them.

1. Assuming a file manifest of training data with species labels, first split the data into training, validation and test splits. This function splits each label proportionally by the given percentages, by default 0.8 training, 0.1 validation, 0.1 Test.

   train, val, test, stats = animl.export_train_val_test(manifest,
                                                         out_dir='path/to/save/data/',
                                                         label_col="species",
                                                         val_size: float = 0.1,
                                                         test_size: float = 0.1,
                                                         random_state: int = 42)
   

2. Set up training configuration file. Specify the paths to the data splits from the previous step. Example configs are available in the configs/ folder.

3. Using the config file, begin training.

   python -m animl.train --config /path/to/config.yaml
   

   Every 10 epochs (or define custom 'checkpoint_frequency'), the model will be checkpointed to the 'experiment_folder' parameter in the config file, and will contain performance metrics for selection.

4. Testing of a model checkpoint can be done with the "test.py" module. Add an 'active_model' parameter to the config file that contains the path of the checkpoint to test. This will produce a confusion matrix of the test dataset as well as a csv containing predicted and ground truth labels for each image.

   python -m animl.test --config /path/to/config.yaml
   

Models

The Conservation Technology Lab has several models available for use. You can use the download function within animl or access them here:

animl.download_model(animl.CLASSIFIER['sdzes_andes_v1'],  out_dir: str = 'models/')

• Southwest United States v3
• Amazon
• Savannah
• Andes
• MegaDetector