FastSAM s | Luxonis

Our new model ZOO works with DepthAI V3. Find out more in our documentation.

0 Likes

Model Details

Model Description

The Fast Segment Anything Model (FastSAM) is a novel, real-time CNN-based solution for the Segment Anything task. This task is designed to segment any object within an image based on various possible user interaction prompts. FastSAM significantly reduces computational demands while maintaining competitive performance, making it a practical choice for a variety of vision tasks. We implement here the s version of the model.

Developed by: Ultralytics
Shared by:
Model type: Computer Vision
License:
Resources for more information:

Training Details

Training Data

The model was trained on only 2% of the dataset. Segment Anything 1 Billion (SA-1B) is a dataset designed for training general-purpose object segmentation models from open world images.

Testing Details

Testing details are only available for the bigger FastSAM x variant.

Technical Specifications

Input/Output Details

Input:
- Name: image
  - Info: NCHW BGR un-normalized image
Output:
- Name: Multiple (please consult NN archive config.json)
  - Info: Unprocessed outputs of a multitude of detections, masks and protos

Model Architecture

Backbone: CSPDarknet53
Head: Anchor-free object segmentation head from YOLOv8 seg model (pruned of concatenation)

Consult the for more information.

Throughput

Model variant: fastsam-s:512x288

• Input shape: [1, 3, 288, 512] • Output shapes: [[1, 6, 36, 64], [1, 6, 18, 32], [1, 6, 9, 16], [1, 32, 36, 64], [1, 32, 18, 32], [1, 32, 9, 16], [1, 32, 72, 128]]

• Params (M): 11.780 • GFLOPs: 8.037

Platform	Precision	Throughput (infs/sec)	Power Consumption (W)
RVC2	FP16	14.15	N/A
RVC4	INT8	491.72	4.25

* Benchmarked with , using 2 threads (and the DSP runtime in balanced mode for RVC4).

* Parameters and FLOPs are obtained from the package.

Quantization

RVC4 version of the model was quantized using a custom dataset. This was created by taking a full 128-image dataset.

Utilization

Models converted for RVC Platforms can be used for inference on OAK devices. DepthAI pipelines are used to define the information flow linking the device, inference model, and the output parser (as defined in model head(s)). Below, we present the most crucial utilization steps for the particular model. Please consult the docs for more information.

Install DAIv3 and depthai-nodes libraries:

pip install depthai
pip install depthai-nodes

Define model:

model_description = dai.NNModelDescription(
    "luxonis/fastsam-s:512x288"
)

nn = pipeline.create(ParsingNeuralNetwork).build(
    <CameraNode>, model_description
)

Inspect model head(s):

FastSAMParser that outputs message (mask of each of the segmented objects).

Get parsed output(s):

while pipeline.isRuning():
    parser_output: SegmentationMask = parser_output_queue.get()

Notes

CLIP textual and visual models are used for text prompting, to determine image segments corresponding to the given text prompt. Both these models are encoders, the difference between them is that textual CLIP is used to encode an input text, whereas visual CLIP is used to encode images (or segments of images). To learn more about various prompting methods, please refer to the .

Example

You can quickly run the model using our script. It automatically downloads the model, creates a DepthAI pipeline, runs the inference, and displays the results using our DepthAI visualizer tool. To try it out, run:

python3 main.py \
    --model luxonis/fastsam-s:512x288 \
    -overlay

A segment anything type of a model.
License	GNU Affero General Public License v3.0 Commercial use
Downloads	296
Tasks	Semantic Segmentation
Model Types	ONNX

Model Variants

Name	Version	Available For	Created At	Deploy
		RVC2, RVC4	10 months ago