## Available data

The test data is organised under subdirectories that match the tested modules in the `ethology` package.

### Test data for `ethology.annotations`

#### `medium_bboxes_dataset_VIA`

A medium size bounding boxes dataset in VIA format.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `VIA_JSON_sample_1.json` | bboxes | VIA | 4440 | 50 | Category ID is a string ("crab"). |
| `VIA_JSON_sample_2.json` | bboxes | VIA | 3977 | 50 | Category ID is a string ("crab").|

#### `medium_bboxes_dataset_COCO`

A medium size bounding boxes dataset in COCO format.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `COCO_JSON_sample_1.json` | bboxes | COCO | 4344 | 100 | Category ID is a number (1). |
| `COCO_JSON_sample_2.json` | bboxes | COCO | 4618 | 100 | Category ID is a number (1).|

#### `small_bboxes_dataset`

A small size bounding boxes dataset in VIA, COCO and VIA-CSV formats.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_VIA.json` | bboxes | VIA | 3 | 3 | One bounding box per frame. The category ID is "1" (string) for all annotations. |
| `small_bboxes_COCO.json` | bboxes | COCO | 3 | 3 | One bounding box per frame. The category ID is 1 (integer) for all annotations. |

#### `small_bboxes_dataset_duplicates`

A small size bounding boxes dataset in VIA, COCO and VIA-CSV formats with duplicate annotations.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_duplicates_VIA.json` | bboxes | VIA | 4 | 3 | Duplicate annotation in frame 0, one annotation per frame in the rest of frames. |
| `small_bboxes_duplicates_COCO.json` | bboxes | COCO | 4 | 3 |  Duplicate annotation in frame 0, one annotation per frame in the rest of frames.|

#### `small_bboxes_dataset_no_categories`

A small size bounding boxes dataset in VIA, COCO and VIA-CSV formats without categories.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_no_cat_VIA.json` | bboxes | VIA | 3 | 3 | No categories or supercategories are specified. This data is still supported |
| `small_bboxes_no_cat_COCO.json` | bboxes | COCO | 3 | 3 |  No categories are specified, so no annotations are exported in COCO format. This data should throw an error. |
| `small_bboxes_null_catID_COCO.json` | bboxes | COCO | 3 | 3 | The category ID in the VIA tool is not an integer, so when exporting the annotations to COCO format the category ID is set to `null`. This data should throw an error. |
| `small_bboxes_no_supercat_VIA.json` | bboxes | VIA | 3 | 3 | A category (`crab`) is specified, but supercategory is set to `" "` (whitespace). This file is supported |
| `small_bboxes_no_supercat_COCO.json` | bboxes | COCO | 3 | 3 |  A category (`crab`) is specified, but no supercategories. This data is supported. |

#### `small_bboxes_dataset_subset`

A subset of the corresponding small bounding boxes datasets. If prefixed with `_plus` it contains some additional annotations. See description of the corresponding dataset for more details. The files with the same name except for the format identifier contain the exact same annotations.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_VIA_subset.json` | bboxes | VIA | 2 | 3 | Same as `small_bboxes_VIA.json` but the last annotation in the last frame has been removed. |
| `small_bboxes_VIA_subset_plus.json` | bboxes | VIA | 3 | 3 | Same as `small_bboxes_VIA_subset.json` but with one additional annotation, different from any of the annotations in `small_bboxes_VIA.json`.|
| `small_bboxes_COCO_subset.json` | bboxes | VIA | 2 | 3 | Same as `small_bboxes_COCO.json` but the last annotation in the last frame has been removed. |
| `small_bboxes_COCO_subset_plus.json` | bboxes | VIA | 3 | 3 | Same as `small_bboxes_COCO_subset.json` but with one additional annotation, different from any of the annotations in `small_bboxes_COCO.json`.|


#### `small_bboxes_image_id_checks`

Small size bounding boxes dataset to check if the image ID based on the alphabetical order of the filenames is assigned correctly by `ethology`.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_image_id_VIA.json` | bboxes | VIA | 3 | 3 | In the `_via_image_id_list` the keys for the three image filenames is not sorted alphabetically. So the image ID assigned based on the order of `_via_image_id_list` is different to the image ID assigned based on the alphabetically sorted list of filenames. |
| `small_bboxes_image_id_COCO.json` | bboxes | COCO | 3 | 3 | The image IDs in the COCO file are different from the image IDs that would be assigned if based on the alphabetically sorted list of filenames. |


#### `small_bboxes_dataset_errors`

Small size bounding boxes datasets with errors to catch.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_non_unique_img_id_COCO.json` | bboxes | COCO | 4 | 4 | Two images have the same image ID and their filename differs only by the extension. This happens if two images have the same number-based filename, because the imageID is extracted using [`parseInt`](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/parseInt) when exporting as COCO in the VIA tool.|


#### `small_bboxes_image_shape`

Small size bounding boxes datasets with image shape data as exported by the VIA tool. The image width and height are stored as "file_attributes" in the VIA file.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `small_bboxes_image_shape_COCO.json` | bboxes | COCO | 3 | 3 | One annotation per image. The image shape is stored as an integer in the exported JSON file. |
| `small_bboxes_image_shape_VIA.json` | bboxes | VIA | 3 | 3 | One annotation per image. The image shape is stored as a string in the exported JSON file. |

#### `sample_external_COCO_files`

COCO files not exported from the VIA tool.

| File name | Type | Format | N annotations | N images | Comments |
| --- | --- | --- | --- | --- | --- |
| `SnapshotSerengetiBboxes_20190903.json` | bboxes |  COCO Camera Traps .json format | 145380 | 78029 with annotations | File obtained from the [Snapshot Serengeti dataset](https://lila.science/datasets/snapshot-serengeti). Image IDs and annotation IDs are strings of numbers and letters. The data includes three categories ("animal", "person" and "group"). |
| `ACTD_1_Terrestrial_group_data_CCT.json` | bboxes |  COCO Camera Traps .json format | 19977 | 17707 images, but only 17659 have annotations | File obtained from the [Australian Camera Trap Data (ACTD) dataset](https://figshare.com/articles/dataset/Australian_Camera_Trap_Data_ACTD_/27177912?file=53674187). No supercategories are specified. Image width and height vary across the dataset. Coordinates are in normalised format (from 0 to 1). The data includes two categories ("animal" and "person"). |