Skip to main content

Crate burn_autogaze

Crate burn_autogaze 

Source

Structs§

AutoGazeCausalLmOutput
AutoGazeClipShape
AutoGazeConfig
AutoGazeDeviceGenerateOutput
AutoGazeDeviceMask
AutoGazeDeviceReadoutRunOutput
AutoGazeDeviceTokens
AutoGazeEmaMetric
AutoGazeEmbedOutput
AutoGazeGazeRatioStats
AutoGazeGazingModel
AutoGazeGenerateOutput
AutoGazeInferenceSequencer
Monotonic sequence gate for asynchronous AutoGaze inference results.
AutoGazeLoadOptions
AutoGazeMaskPlanStats
AutoGazePatchDiffConfig
AutoGazePatchDiffDeviceMask
AutoGazePipeline
AutoGazePipelineOptions
AutoGazePipelinePacket
AutoGazePreparedRun
AutoGazePsnrStats
AutoGazeReadoutRunOutput
Selected fixation points from a prepared AutoGaze run.
AutoGazeReadoutStats
AutoGazeRealtimePolicy
Frontend admission policy for realtime AutoGaze streams.
AutoGazeRgbaClip
AutoGazeRgbaClipShape
AutoGazeRgbaFrameClip
AutoGazeRgbaFrameQueue
AutoGazeRgbaVisualizationBuffers
Reusable RGBA work buffers for allocation-stable CPU visualization.
AutoGazeRgbaVisualizationOptions
AutoGazeScaleTokenLayout
AutoGazeScaleTokenMask
AutoGazeSparseUpdatePlan
AutoGazeStreamingCache
Stateful generation cache for advancing an AutoGaze video stream one frame at a time.
AutoGazeTensorClip
AutoGazeTensorClipShape
AutoGazeTensorPipeline
AutoGazeTensorPipelineConfig
AutoGazeTensorVisualization
AutoGazeTensorVisualizationOptions
AutoGazeTensorVisualizationPanels
AutoGazeTensorVisualizationState
AutoGazeTile
AutoGazeTileLayout
AutoGazeTraceRunOutput
AutoGazeTraceStore
AutoGazeVisualization
AutoGazeVisualizationPanels
AutoGazeVisualizationPanelsView
AutoGazeVisualizationState
Connector
ConnectorConfig
Conv3dBlockForStreaming
FixationBounds
FixationPixelRect
FixationPoint
FixationSet
FnOutputNode
FrameFixationTrace
GazeDecoderConfig
GazeModelConfig
ImagePyramidLevel
ImagePyramidMask
ImagePyramidMaskOptions
ImagePyramidTokens
NativeAutoGazeModel
RgbaClipInput
ShallowVideoConvNet
SparseImagePyramidTokens
SparseReadoutGrid
Downstream image-token grid used to project AutoGaze regions into sparse readout tokens.
SparseReadoutOptions
Projection options for converting AutoGaze fixations into sparse image readout tokens.
SparseReadoutRect
Normalized image-space rectangle selected for sparse downstream readout.
SparseVideoPatchGeometry
Video/tubelet/patch geometry for downstream sparse video patchifiers.
SparseVideoReadoutGrid
Sparse video-token grid used to project per-frame image readout into downstream tubelet or temporal-token layouts.
SparseVideoReadoutOptions
Projection options for converting per-frame image readout into sparse video token indices.
SparseVideoReadoutProjection
Complete sparse-video projection settings for adapting AutoGaze image readout into a downstream video-token grid.
TensorClipInput
VecOutputNode
VisionModelConfig

Enums§

AutoGazeDecodeStrategy
Greedy decoder readout strategy for streaming AutoGaze generation.
AutoGazeInferenceMode
AutoGazeMaskGeometryMode
AutoGazeMaskVisualizationMode
AutoGazeSparseMaskSource
AutoGazeTaskLossOption
AutoGazeTensorInterframePath
AutoGazeVisualizationMode

Constants§

AUTO_GAZE_IMAGE_MEAN
AUTO_GAZE_IMAGE_STD
AUTO_GAZE_PROCESSOR_SHORT_EDGE
AUTO_GAZE_RESCALE_FACTOR
DEFAULT_BLEND_ALPHA
Default alpha for readable white-mask overlays.
DEFAULT_KEYFRAME_DURATION
Default keyframe interval for interframe visualization.
DEFAULT_MAX_IN_FLIGHT
Default maximum number of realtime AutoGaze inference tasks in flight.
DEFAULT_METRIC_EMA_ALPHA
Default exponential moving average coefficient used by AutoGaze runtime metrics.
DEFAULT_MODEL_GENERATION_BUDGET
Wrapper sentinel for keeping the model’s configured generation budget.
DEFAULT_PATCH_DIFF_GRID_SIZE
DEFAULT_PATCH_DIFF_THRESHOLD
DEFAULT_REALTIME_FRAMES_PER_CLIP
Default number of frames kept in realtime input windows.
DEFAULT_REALTIME_TOP_K
Default number of displayed trace slots for realtime resize-mode streams.
DEFAULT_TENSOR_FULL_FRAME_UPDATE_MIN_RATIO
DEFAULT_TENSOR_SPARSE_UPDATE_MAX_RATIO
DEFAULT_TENSOR_SPARSE_UPDATE_MAX_RECTS
DEFAULT_TILED_FRAMES_PER_CLIP
Default number of frames kept in tiled input windows.
DEFAULT_TILED_MAX_GAZE_TOKENS
Default per-frame generated-token cap for tiled inspection modes.
DEFAULT_TILED_TILE_BATCH_SIZE
Default tile batch size for tiled AutoGaze embedding and tracing.
DEFAULT_TILED_TOP_K
Default displayed trace slots per tile for tiled inspection modes.

Traits§

AutoGazeInputNode
AutoGazeOutputNode
AutoGazeTeacher

Functions§

apply_image_mask
batched_video_readout_tokens_to_coord_tensor
Convert per-batch sparse video-token indices directly into a flattened Burn coordinate tensor with shape [rows, 4].
batched_video_readout_tokens_to_coords
Convert per-batch sparse video-token indices into flattened sparse patchification coordinates.
copy_sparse_update_rgba
Copy source RGBA pixels for sparse update rectangles into a persistent output frame.
copy_sparse_update_tensor
ema_metric
fixation_alpha_mask
fixation_cell_rects
fixation_deduplicated_cell_rects
fixation_deduplicated_sparse_update_plan
Build native-scale sparse update rectangles after dropping fully covered cells.
fixation_effective_alpha_mask
Build a projected sparse footprint from multi-scale gaze tokens.
fixation_effective_cell_rects
fixation_effective_scale_mask_rgba
Colorize the same projected cells used by fixation_effective_alpha_mask.
fixation_effective_scale_mask_rgba_into
fixation_effective_sparse_update_plan
Build sparse update rectangles projected onto the finest active gaze grid.
fixation_image_mask_only_rgba
fixation_image_mask_only_rgba_into
fixation_image_mask_tensor
fixation_image_overlay_mask_rgba
fixation_image_overlay_mask_rgba_into
fixation_mask_rgba
fixation_mask_rgba_into
fixation_points_to_readout_rects
Convert AutoGaze fixation points into normalized readout rectangles.
fixation_points_to_readout_tokens
Convert AutoGaze fixation points directly into sparse image readout token indices.
fixation_rect_union_pixel_count
fixation_scale_mask_rgba
fixation_scale_mask_rgba_into
fixation_scale_rows_mask_rgba
fixation_scale_rows_mask_rgba_into
fixation_sparse_update_plan
Build the native-scale sparse update rectangles used by the default AutoGaze visualization.
format_fps
format_gaze_ratio_percent
format_psnr_db
fps_from_millis
frame_fixation_masks_tensor
frame_readout_rects_to_video_coord_tensor
Project per-frame normalized readout rectangles directly into a Burn sparse video-coordinate tensor with shape [rows, 4].
frame_readout_rects_to_video_coords
Project per-frame normalized readout rectangles into sparse video-token coordinates for downstream sparse patchification kernels.
frame_readout_rects_to_video_tokens
Project per-frame normalized readout rectangles into a sparse video-token grid.
frame_readout_tokens_to_video_coord_tensor
Project per-frame image readout token indices directly into a Burn sparse video-coordinate tensor with shape [rows, 4].
frame_readout_tokens_to_video_coords
Project per-frame image readout token indices into sparse video-token coordinates for downstream sparse patchification kernels.
frame_readout_tokens_to_video_tokens
Project per-frame image readout token indices into a sparse video-token grid.
generated_frame_readout_rects
Convert one frame from generated AutoGaze output into normalized readout rectangles.
generated_frame_readout_tokens
Convert one frame from generated AutoGaze output into sparse image readout token indices.
generated_to_frame_readout_rects
Convert every frame from generated AutoGaze output into normalized readout rectangles.
generated_to_frame_readout_tokens
Convert every frame from generated AutoGaze output into sparse image readout token indices.
generated_to_video_readout_coord_tensor
Convert generated AutoGaze output directly into a Burn sparse-video coordinate tensor with shape [rows, 4].
generated_to_video_readout_coords
Convert generated AutoGaze output directly into downstream sparse-video coordinates.
generated_to_video_readout_tokens
Convert generated AutoGaze output directly into downstream sparse video-token indices.
image_pyramid_masks
last_rgba_frame
normalized_rgb_clip_to_unit_rgba_tensor
patch_diff_device_mask_async
patch_diff_points_to_traces
patch_diff_readout_points
patch_diff_readout_points_async
patch_diff_scores
prepare_rgba_clip_for_trace
readout_rects_to_tokens
Project normalized readout rectangles onto a sparse image-token grid.
resize_dimensions_preserving_aspect
resize_rgba_frame_to_dimensions
resize_video_shortest_edge
rgba_clip_to_inference_tensor
rgba_clip_to_processor_tensor
rgba_clip_to_tensor
sanitize_gaze_ratio
scale_token_layouts
should_use_streaming_cache
sparsify_image_pyramid_tokens
task_loss_requirement_from_l1_db
task_loss_requirement_to_l1_db
tokenize_masked_image_pyramid
trace_frame_readout_rects
Convert one frame of an AutoGaze trace into normalized readout rectangles.
trace_frame_readout_tokens
Convert one frame of an AutoGaze trace into sparse image readout token indices.
trace_to_frame_readout_rects
Convert every frame in an AutoGaze trace into normalized readout rectangles.
trace_to_frame_readout_tokens
Convert every frame in an AutoGaze trace into sparse image readout token indices.
trace_to_video_readout_coord_tensor
Convert an AutoGaze trace directly into a Burn sparse-video coordinate tensor with shape [rows, 4].
trace_to_video_readout_coords
Convert an AutoGaze trace directly into downstream sparse-video coordinates.
trace_to_video_readout_tokens
Convert an AutoGaze trace directly into downstream sparse video-token indices.
video_frame_tensor
video_readout_coords_to_tensor
Convert sparse patchification coordinate rows into a Burn int tensor with shape [rows, 4].
video_readout_tokens_to_coord_tensor
Convert sparse video-token indices directly into a Burn coordinate tensor with shape [rows, 4].
video_readout_tokens_to_coords
Convert sparse video-token indices into [batch, temporal, row, col] coordinates for downstream sparse patchification kernels.
visualize_fixations_rgba

Type Aliases§

NdArrayAutoGazeModel
NdArrayAutoGazePipeline
WebGpuAutoGazeModel
WebGpuAutoGazePipeline
WgpuAutoGazeModel
WgpuAutoGazePipeline