In this format, `<class-index>` is the index of the class for the object,`<x><y><width><height>` are coordinates of boudning box, and `<px1> <py1> <px2> <py2> ... <pxn> <pyn>` are the pixel coordinates of the keypoints. The coordinates are separated by spaces.
In this format, `<class-index>` is the index of the class for the object,`<x><y><width><height>` are coordinates of boudning box, and `<px1> <py1> <px2> <py2> ... <pxn> <pyn>` are the pixel coordinates of the keypoints. The coordinates are separated by spaces.
** Dataset file format **
### Dataset YAML format
The Ultralytics framework uses a YAML file format to define the dataset and model configuration for training Detection Models. Here is an example of the YAML format used for defining a detection dataset:
The Ultralytics framework uses a YAML file format to define the dataset and model configuration for training Detection Models. Here is an example of the YAML format used for defining a detection dataset:
```yaml
```yaml
train: <path-to-training-images>
# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
val: <path-to-validation-images>
path: ../datasets/coco8-pose # dataset root dir
train: images/train # train images (relative to 'path') 4 images
nc: <number-of-classes>
val: images/val # val images (relative to 'path') 4 images
names: [<class-1>, <class-2>, ..., <class-n>]
test: # test images (optional)
# Keypoints
# Keypoints
kpt_shape: [num_kpts, dim] # number of keypoints, number of dims (2 for x,y or 3 for x,y,visible)
kpt_shape: [17, 3] # number of keypoints, number of dims (2 for x,y or 3 for x,y,visible)
The `train` and `val` fields specify the paths to the directories containing the training and validation images, respectively.
The `nc` field specifies the number of object classes in the dataset.
The `names` field is a list of the names of the object classes. The order of the names should match the order of the object class indices in the YOLO dataset files.
NOTE: Either `nc` or `names` must be defined. Defining both are not mandatory
Alternatively, you can directly define class names like this:
```yaml
# Classes dictionary
names:
names:
0: person
0: person
1: bicycle
```
```
(Optional) if the points are symmetric then need flip_idx, like left-right side of human or face.
The `train` and `val` fields specify the paths to the directories containing the training and validation images, respectively.
For example let's say there're five keypoints of facial landmark: [left eye, right eye, nose, left point of mouth, right point of mouse], and the original index is [0, 1, 2, 3, 4], then flip_idx is [1, 0, 2, 4, 3].(just exchange the left-right index, i.e 0-1 and 3-4, and do not modify others like nose in this example)
** Example **
```yaml
train: data/train/
val: data/val/
nc: 2
`names` is a dictionary of class names. The order of the names should match the order of the object class indices in the YOLO dataset files.
names: ['person', 'car']
# Keypoints
(Optional) if the points are symmetric then need flip_idx, like left-right side of human or face.
kpt_shape: [17, 3] # number of keypoints, number of dims (2 for x,y or 3 for x,y,visible)
For example if we assume five keypoints of facial landmark: [left eye, right eye, nose, left mouth, right mouth], and the original index is [0, 1, 2, 3, 4], then flip_idx is [1, 0, 2, 4, 3] (just exchange the left-right index, i.e 0-1 and 3-4, and do not modify others like nose in this example).
@ -37,44 +37,31 @@ Here is an example of the YOLO dataset format for a single image with two object
Note: The length of each row does not have to be equal.
Note: The length of each row does not have to be equal.
** Dataset file format **
### Dataset YAML format
The Ultralytics framework uses a YAML file format to define the dataset and model configuration for training Detection Models. Here is an example of the YAML format used for defining a detection dataset:
The Ultralytics framework uses a YAML file format to define the dataset and model configuration for training Detection Models. Here is an example of the YAML format used for defining a detection dataset:
```yaml
```yaml
train: <path-to-training-images>
# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
val: <path-to-validation-images>
path: ../datasets/coco8-seg # dataset root dir
train: images/train # train images (relative to 'path') 4 images
val: images/val # val images (relative to 'path') 4 images
test: # test images (optional)
nc: <number-of-classes>
# Classes (80 COCO classes)
names: [<class-1>, <class-2>, ..., <class-n>]
```
The `train` and `val` fields specify the paths to the directories containing the training and validation images, respectively.
The `nc` field specifies the number of object classes in the dataset.
The `names` field is a list of the names of the object classes. The order of the names should match the order of the object class indices in the YOLO dataset files.
NOTE: Either `nc` or `names` must be defined. Defining both are not mandatory.
Alternatively, you can directly define class names like this:
```yaml
names:
names:
0: person
0: person
1: bicycle
1: bicycle
2: car
...
77: teddy bear
78: hair drier
79: toothbrush
```
```
** Example **
The `train` and `val` fields specify the paths to the directories containing the training and validation images, respectively.
```yaml
train: data/train/
val: data/val/
nc: 2
`names` is a dictionary of class names. The order of the names should match the order of the object class indices in the YOLO dataset files.
@ -12,7 +12,7 @@ The Fast Segment Anything Model (FastSAM) is a novel, real-time CNN-based soluti
## Overview
## Overview
FastSAM is designed to address the limitations of the Segment Anything Model (SAM), a heavy Transformer model with substantial computational resource requirements. The FastSAM decouples the segment anything task into two sequential stages: all-instance segmentation and prompt-guided selection. The first stage uses a Convolutional Neural Network (CNN)-based detector to produce the segmentation masks of all instances in the image. In the second stage, it outputs the region-of-interest corresponding to the prompt.
FastSAM is designed to address the limitations of the [Segment Anything Model (SAM)](sam.md), a heavy Transformer model with substantial computational resource requirements. The FastSAM decouples the segment anything task into two sequential stages: all-instance segmentation and prompt-guided selection. The first stage uses [YOLOv8-seg](../tasks/segment.md) to produce the segmentation masks of all instances in the image. In the second stage, it outputs the region-of-interest corresponding to the prompt.
## Key Features
## Key Features
@ -22,9 +22,9 @@ FastSAM is designed to address the limitations of the Segment Anything Model (SA
3. **Prompt-guided Segmentation:** FastSAM can segment any object within an image guided by various possible user interaction prompts, providing flexibility and adaptability in different scenarios.
3. **Prompt-guided Segmentation:** FastSAM can segment any object within an image guided by various possible user interaction prompts, providing flexibility and adaptability in different scenarios.
4. **Based on YOLOv8-seg:** FastSAM is based on YOLOv8-seg, an object detector equipped with an instance segmentation branch. This allows it to effectively produce the segmentation masks of all instances in an image.
4. **Based on YOLOv8-seg:** FastSAM is based on [YOLOv8-seg](../tasks/segment.md), an object detector equipped with an instance segmentation branch. This allows it to effectively produce the segmentation masks of all instances in an image.
5. **Competitive Results on Benchmarks:** On the object proposal task on MS COCO, FastSAM achieves high scores at a significantly faster speed than SAM on a single NVIDIA RTX 3090, demonstrating its efficiency and capability.
5. **Competitive Results on Benchmarks:** On the object proposal task on MS COCO, FastSAM achieves high scores at a significantly faster speed than [SAM](sam.md) on a single NVIDIA RTX 3090, demonstrating its efficiency and capability.
6. **Practical Applications:** The proposed approach provides a new, practical solution for a large number of vision tasks at a really high speed, tens or hundreds of times faster than current methods.
6. **Practical Applications:** The proposed approach provides a new, practical solution for a large number of vision tasks at a really high speed, tens or hundreds of times faster than current methods.
@ -32,7 +32,7 @@ FastSAM is designed to address the limitations of the Segment Anything Model (SA
## Usage
## Usage
FastSAM is not yet available directly via the `ultralytics` package, but it is available directly from the [https://github.com/CASIA-IVA-Lab/FastSAM](https://github.com/CASIA-IVA-Lab/FastSAM) repository. Here is a brief overview of the typical steps you might take to use FastSAM:
FastSAM is not yet available within the [`ultralytics` package](../quickstart.md), but it is available directly from the [https://github.com/CASIA-IVA-Lab/FastSAM](https://github.com/CASIA-IVA-Lab/FastSAM) repository. Here is a brief overview of the typical steps you might take to use FastSAM:
Glenn Jocher
1 year ago
committed by
GitHub