Transformers documentation

VidEoMT

Transformers

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v5.3.0).

Join the Hugging Face community

and get access to the augmented documentation experience

Collaborate on models, datasets and Spaces

Faster examples with accelerated inference

Switch between documentation themes

to get started

This model was released on 2026-02-19 and added to Hugging Face Transformers on 2026-03-25.

VidEoMT

Overview

The VidEoMT model was proposed in Your ViT is Secretly Also a Video Segmentation Model by Narges Norouzi, Idil Esen Zulfikar, Niccolò Cavagnero, Tommie Kerssies, Bastian Leibe, Gijs Dubbelman, Daan de Geus. Video Encoder-only Mask Transformer (VidEoMT) is a lightweight encoder-only model for online video segmentation built on a plain Vision Transformer (ViT). It is a minimal extension of EoMT to video which performs both spatial and temporal reasoning within the ViT encoder, without relying on dedicated tracking modules or heavy task-specific heads.

The abstract from the paper is the following:

Existing online video segmentation models typically combine a per-frame segmenter with complex specialized tracking modules. While effective, these modules introduce significant architectural complexity and computational overhead. Recent studies suggest that plain Vision Transformer (ViT) encoders, when scaled with sufficient capacity and large-scale pre-training, can conduct accurate image segmentation without requiring specialized modules. Motivated by this observation, we propose the Video Encoder-only Mask Transformer (VidEoMT), a simple encoder-only video segmentation model that eliminates the need for dedicated tracking modules. To enable temporal modeling in an encoder-only ViT, VidEoMT introduces a lightweight query propagation mechanism that carries information across frames by reusing queries from the previous frame. To balance this with adaptability to new content, it employs a query fusion strategy that combines the propagated queries with a set of temporally-agnostic learned queries. As a result, VidEoMT attains the benefits of a tracker without added complexity, achieving competitive accuracy while being 5x—10x faster, running at up to 160 FPS with a ViT-L backbone.

Tips:

VidEoMT currently only supports a DINOv2 backbone (with register tokens). Available model sizes are ViT-S, ViT-B, and ViT-L.
The model accepts video input as a 5D tensor of shape (batch_size, num_frames, 3, height, width).
VidEoMT supports three video segmentation tasks: instance, semantic, and panoptic segmentation, each with a dedicated post-processing method on the video processor.

This model was contributed by nielsr. The original code can be found here.

Architecture Info

VidEoMT builds on EoMT, which repurposes a plain DINOv2-pretrained Vision Transformer with register tokens as a segmentation model. EoMT introduces learned object queries and a lightweight mask prediction head directly inside the ViT encoder, eliminating the need for task-specific decoders.

VidEoMT extends this to video with two key additions:

Query propagation: object queries from the previous frame are carried forward to the next frame through a linear projection (query_updater), enabling temporal reasoning without a dedicated tracker.
Query fusion: the propagated queries are added to a set of temporally-agnostic learned queries, allowing the model to adapt to new objects appearing in the video.

The early encoder layers process all frames independently (in parallel), while the final blocks operate per-frame with the fused queries, producing per-frame mask and class predictions.

Usage Examples

Use the Hugging Face implementation of VidEoMT for inference with pre-trained models. The examples below reuse the public tue-mps/videomt-dinov2-small-ytvis2019 checkpoint to demonstrate video instance, semantic, and panoptic post-processing on a sample video.

Video Instance Segmentation

import matplotlib.pyplot as plt
import numpy as np
import torch

from transformers import AutoModelForUniversalSegmentation, AutoVideoProcessor
from transformers.video_utils import load_video


model_id = "tue-mps/videomt-dinov2-small-ytvis2019"
processor = AutoVideoProcessor.from_pretrained(model_id)
model = AutoModelForUniversalSegmentation.from_pretrained(model_id, device_map="auto")

video_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/videos/pexels-allan-mas-5362370.mp4"
# Sample 8 frames to keep the example lightweight.
video_frames, _ = load_video(video_url, num_frames=8)

inputs = processor(videos=[video_frames], return_tensors="pt").to(model.device)

with torch.inference_mode():
    outputs = model(**inputs)

original_height, original_width = video_frames[0].shape[:2]
target_sizes = [(original_height, original_width)] * len(video_frames)

results = processor.post_process_instance_segmentation(
    outputs,
    target_sizes=target_sizes,
)

fig, axes = plt.subplots(2, 4, figsize=(16, 8))
for idx, (ax, frame, result) in enumerate(zip(axes.flatten(), video_frames, results)):
    ax.imshow(frame)
    seg = result["segmentation"].cpu().numpy()
    masked = np.ma.masked_where(seg == -1, seg)
    ax.imshow(masked, alpha=0.6, cmap="tab20")
    ax.set_title(f"Frame {idx}")
    ax.axis("off")
plt.suptitle("Video Instance Segmentation")
plt.tight_layout()
plt.show()

Video Semantic Segmentation

import matplotlib.pyplot as plt
import torch

from transformers import AutoModelForUniversalSegmentation, AutoVideoProcessor
from transformers.video_utils import load_video


model_id = "tue-mps/videomt-dinov2-small-ytvis2019"
processor = AutoVideoProcessor.from_pretrained(model_id)
model = AutoModelForUniversalSegmentation.from_pretrained(model_id, device_map="auto")

video_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/videos/pexels-allan-mas-5362370.mp4"
# Sample 8 frames to keep the example lightweight.
video_frames, _ = load_video(video_url, num_frames=8)

inputs = processor(videos=[video_frames], return_tensors="pt").to(model.device)

with torch.inference_mode():
    outputs = model(**inputs)

original_height, original_width = video_frames[0].shape[:2]
target_sizes = [(original_height, original_width)] * len(video_frames)

preds = processor.post_process_semantic_segmentation(
    outputs,
    target_sizes=target_sizes,
)

fig, axes = plt.subplots(2, 4, figsize=(16, 8))
for idx, (ax, frame, seg_map) in enumerate(zip(axes.flatten(), video_frames, preds)):
    ax.imshow(frame)
    ax.imshow(seg_map.cpu().numpy(), alpha=0.6, cmap="tab20")
    ax.set_title(f"Frame {idx}")
    ax.axis("off")
plt.suptitle("Video Semantic Segmentation")
plt.tight_layout()
plt.show()

Video Panoptic Segmentation

import matplotlib.pyplot as plt
import numpy as np
import torch

from transformers import AutoModelForUniversalSegmentation, AutoVideoProcessor
from transformers.video_utils import load_video


model_id = "tue-mps/videomt-dinov2-small-ytvis2019"
processor = AutoVideoProcessor.from_pretrained(model_id)
model = AutoModelForUniversalSegmentation.from_pretrained(model_id, device_map="auto")

video_url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/videos/pexels-allan-mas-5362370.mp4"
# Sample 8 frames to keep the example lightweight.
video_frames, _ = load_video(video_url, num_frames=8)

inputs = processor(videos=[video_frames], return_tensors="pt").to(model.device)

with torch.inference_mode():
    outputs = model(**inputs)

original_height, original_width = video_frames[0].shape[:2]
target_sizes = [(original_height, original_width)] * len(video_frames)

results = processor.post_process_panoptic_segmentation(
    outputs,
    target_sizes=target_sizes,
)

fig, axes = plt.subplots(2, 4, figsize=(16, 8))
for idx, (ax, frame, result) in enumerate(zip(axes.flatten(), video_frames, results)):
    ax.imshow(frame)
    seg = result["segmentation"].cpu().numpy()
    masked = np.ma.masked_where(seg == -1, seg)
    ax.imshow(masked, alpha=0.6, cmap="tab20")
    ax.set_title(f"Frame {idx}")
    ax.axis("off")
plt.suptitle("Video Panoptic Segmentation")
plt.tight_layout()
plt.show()

VideomtVideoProcessor

class transformers.VideomtVideoProcessor

< source >

( **kwargs: typing_extensions.Unpack[transformers.processing_utils.VideosKwargs] )

post_process_semantic_segmentation

< source >

( outputs target_sizes: list ) → list[torch.Tensor]

Parameters

outputs (VideomtForUniversalSegmentationOutput) — Raw outputs of the model.
target_sizes (list[tuple[int, int]]) — List of (height, width) tuples corresponding to the requested final size of each prediction. Length should match the number of frames in the output.

Returns

list[torch.Tensor]

A list of tensors, each of shape (height, width), where each value is the predicted class index for the corresponding pixel.

Converts the output of VideomtForUniversalSegmentation into semantic segmentation predictions.

post_process_instance_segmentation

< source >

( outputs target_sizes: list threshold: float = 0.5 ) → list[dict]

Parameters

outputs (VideomtForUniversalSegmentationOutput) — Raw outputs of the model.
target_sizes (list[tuple[int, int]]) — List of (height, width) tuples corresponding to the requested final size of each prediction. Length should match the number of frames in the output.
threshold (float, optional, defaults to 0.5) — Minimum combined score to keep an instance.

Returns

list[dict]

A list of dicts (one per frame), each containing:

"segmentation" — A torch.Tensor of shape (height, width) with instance IDs (or -1 for background).
"segments_info" — A list of dicts with "id", "label_id", and "score" for each instance.

Converts the output of VideomtForUniversalSegmentation into instance segmentation predictions.

post_process_panoptic_segmentation

< source >

( outputs target_sizes: list threshold: float = 0.8 mask_threshold: float = 0.5 overlap_mask_area_threshold: float = 0.8 label_ids_to_fuse: set[int] | None = None ) → list[dict]

Parameters

outputs (VideomtForUniversalSegmentationOutput) — Raw outputs of the model.
target_sizes (list[tuple[int, int]]) — List of (height, width) tuples corresponding to the requested final size of each prediction. Length should match the number of frames in the output.
threshold (float, optional, defaults to 0.8) — Minimum score to keep a predicted segment.
mask_threshold (float, optional, defaults to 0.5) — Threshold for binarizing mask probabilities.
overlap_mask_area_threshold (float, optional, defaults to 0.8) — Overlap threshold to merge masks into a single segment.
label_ids_to_fuse (set[int], optional) — Label IDs that should be fused across disconnected regions.

Returns

list[dict]

A list of dicts (one per frame), each containing:

"segmentation" — A torch.Tensor of shape (height, width) with segment IDs (or -1 for background).
"segments_info" — A list of dicts with "id", "label_id", and "score" for each segment.

Converts the output of VideomtForUniversalSegmentation into panoptic segmentation predictions.

VideomtConfig

class transformers.VideomtConfig

< source >

( transformers_version: str | None = None architectures: list[str] | None = None output_hidden_states: bool | None = False return_dict: bool | None = True dtype: typing.Union[str, ForwardRef('torch.dtype'), NoneType] = None chunk_size_feed_forward: int = 0 is_encoder_decoder: bool = False id2label: dict[int, str] | dict[str, str] | None = None label2id: dict[str, int] | dict[str, str] | None = None problem_type: typing.Optional[typing.Literal['regression', 'single_label_classification', 'multi_label_classification']] = None hidden_size: int = 1024 num_hidden_layers: int = 24 num_attention_heads: int = 16 hidden_act: str = 'gelu' hidden_dropout_prob: float = 0.0 initializer_range: float = 0.02 layer_norm_eps: float = 1e-06 image_size: int | list[int] | tuple[int, int] = 640 patch_size: int | list[int] | tuple[int, int] = 16 num_channels: int = 3 mlp_ratio: int = 4 layerscale_value: float = 1.0 drop_path_rate: float = 0.0 num_upscale_blocks: int = 2 attention_dropout: float | int = 0.0 use_swiglu_ffn: bool = False num_blocks: int = 4 no_object_weight: float = 0.1 class_weight: float = 2.0 mask_weight: float = 5.0 dice_weight: float = 5.0 train_num_points: int = 12544 oversample_ratio: float = 3.0 importance_sample_ratio: float = 0.75 num_queries: int = 200 num_register_tokens: int = 4 )

Parameters

hidden_size (int, optional, defaults to 1024) — Dimension of the hidden representations.
num_hidden_layers (int, optional, defaults to 24) — Number of hidden layers in the Transformer decoder.
num_attention_heads (int, optional, defaults to 16) — Number of attention heads for each attention layer in the Transformer decoder.
hidden_act (str, optional, defaults to gelu) — The non-linear activation function (function or string) in the decoder. For example, "gelu", "relu", "silu", etc.
hidden_dropout_prob (float, optional, defaults to 0.0) — The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
layer_norm_eps (float, optional, defaults to 1e-06) — The epsilon used by the layer normalization layers.
image_size (Union[int, list[int], tuple[int, int]], optional, defaults to 640) — The size (resolution) of each image.
patch_size (Union[int, list[int], tuple[int, int]], optional, defaults to 16) — The size (resolution) of each patch.
num_channels (int, optional, defaults to 3) — The number of input channels.
mlp_ratio (int, optional, defaults to 4) — Ratio of the MLP hidden dim to the embedding dim.
layerscale_value (float, optional, defaults to 1.0) — Initial value for the LayerScale parameter.
drop_path_rate (float, optional, defaults to 0.0) — Drop path rate for the patch fusion.
num_upscale_blocks (int, optional, defaults to 2) — Number of upsampling blocks used in the decoder or segmentation head.
attention_dropout (Union[float, int], optional, defaults to 0.0) — The dropout ratio for the attention probabilities.
use_swiglu_ffn (bool, optional, defaults to False) — Whether to use the SwiGLU feedforward neural network.
num_blocks (int, optional, defaults to 4) — Number of feature blocks or stages in the architecture.
no_object_weight (float, optional, defaults to 0.1) — Loss weight for the ‘no object’ class in panoptic/instance segmentation.
class_weight (float, optional, defaults to 2.0) — Loss weight for classification targets.
mask_weight (float, optional, defaults to 5.0) — Loss weight for mask prediction.
dice_weight (float, optional, defaults to 5.0) — Relative weight of the dice loss in the panoptic segmentation loss.
train_num_points (int, optional, defaults to 12544) — Number of points to sample for mask loss computation during training.
oversample_ratio (float, optional, defaults to 3.0) — Oversampling ratio used in point sampling for mask training.
importance_sample_ratio (float, optional, defaults to 0.75) — Ratio of points to sample based on importance during training.
num_queries (int, optional, defaults to 200) — Number of object queries in the Transformer.
num_register_tokens (int, optional, defaults to 4) — Number of learnable register tokens added to the transformer input.

This is the configuration class to store the configuration of a VideomtModel. It is used to instantiate a Videomt model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the tue-mps/coco_panoptic_videomt_large_640

Configuration objects inherit from PreTrainedConfig and can be used to control the model outputs. Read the documentation from PreTrainedConfig for more information.

Example:

>>> from transformers import VideomtConfig, VideomtForUniversalSegmentation

>>> # Initialize configuration
>>> config = VideomtConfig()

>>> # Initialize model
>>> model = VideomtForUniversalSegmentation(config)

>>> # Access config
>>> config = model.config

VideomtPreTrainedModel

class transformers.VideomtPreTrainedModel

< source >

( config: PreTrainedConfig *inputs **kwargs )

Parameters

config (PreTrainedConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

_forward_unimplemented

< source >

( *input: typing.Any )

Define the computation performed at every call.

Should be overridden by all subclasses.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

VideomtForUniversalSegmentation

class transformers.VideomtForUniversalSegmentation

< source >

( config: VideomtConfig )

Parameters

config (VideomtConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The Videomt Model with head on top for instance/semantic/panoptic segmentation.

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward

< source >

( pixel_values_videos: torch.Tensor | None = None mask_labels: list[torch.Tensor] | None = None class_labels: list[torch.Tensor] | None = None patch_offsets: list[torch.Tensor] | None = None **kwargs: typing_extensions.Unpack[transformers.utils.generic.TransformersKwargs] ) → VideomtForUniversalSegmentationOutput or tuple(torch.FloatTensor)

Parameters

pixel_values_videos (torch.Tensor, optional) — Video inputs of shape (batch_size, num_frames, num_channels, height, width).
mask_labels (list[torch.Tensor], optional) — Not supported for 5D video inputs.
class_labels (list[torch.LongTensor], optional) — Not supported for 5D video inputs.
patch_offsets (list[torch.Tensor], optional) — Unused for video inputs and only kept for modular compatibility.

Returns

VideomtForUniversalSegmentationOutput or tuple(torch.FloatTensor)

A VideomtForUniversalSegmentationOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VideomtConfig) and inputs.

The VideomtForUniversalSegmentation forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

loss (torch.Tensor, optional) — The computed loss, returned when labels are present.
class_queries_logits (torch.FloatTensor, optional, defaults to None) — A tensor of shape (batch_size, num_queries, num_labels + 1) representing the proposed classes for each query. Note the + 1 is needed because we incorporate the null class.
masks_queries_logits (torch.FloatTensor, optional, defaults to None) — A tensor of shape (batch_size, num_queries, height, width) representing the proposed masks for each query.
last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Last hidden states (final feature map) of the last layer.
hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each stage) of shape (batch_size, sequence_length, hidden_size). Hidden-states all layers of the model.
attentions (tuple(tuple(torch.FloatTensor)), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of tuple(torch.FloatTensor) (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length). Self and Cross Attentions weights from transformer decoder.

Update on GitHub

←UVDoc Vision Transformer (ViT)→