i.sam2.opencl segments a 3-band imagery group with the Segment Anything Model 2 (SAM 2) and writes the segments as a clean vector area map. The whole image path of the network, Hiera image encoder, feature pyramid neck, prompt encoder and mask decoder, runs on an OpenCL device (GPU or CPU), with weights read directly from the original PyTorch checkpoint file: neither Python nor PyTorch is needed at run time.
The module implements SAM 2's automatic mask generation: the model is prompted with a regular grid of points, and the masks it returns are kept according to their predicted quality and stability, de-duplicated and cleaned. The overlapping masks are then turned into a non-overlapping, topologically clean vector map with one attribute row per segment. SAM 2's video memory (memory encoder and memory attention) is not used for single images and is not run.
The group (or its subgroup) must contain exactly three raster maps, which are used as the red, green and blue channels in the order they are listed in the group (see i.group). Each band is stretched linearly between its lower and upper percentile to the 0-255 range SAM 2 was trained on. Cells that are NULL in any band are not segmented. Processing is done in the current computational region.
The checkpoint is one of the original model files published with SAM 2.1 or SAM 2; the model is detected from the tensor shapes:
The SAM 2 files of the same models (sam2_hiera_large.pt etc., in
https://dl.fbaipublicfiles.com/segment_anything_2/072824/) are
supported too. Checkpoints of the original SAM (sam_vit_h_4b8939.pth
etc.) are refused: they belong to i.sam.opencl.
The weights are too large to be installed with the module. Without a
checkpoint file, the module looks for one in the weights folder:
weights_dir, by default the central folder $SAM_MODELS_DIR/sam2, or
~/models/sam/sam2 when the environment variable SAM_MODELS_DIR is not
set. If the model is not there, its SAM 2.1 checkpoint is downloaded into
that folder (created if needed) on first use and reused afterwards: the
model chosen with model, or Hiera-L when the folder holds no
checkpoint at all. The download is checked against the published size,
resumes where an interrupted one stopped, and waits for another run
downloading the same file. The -n flag forbids downloading: a missing
checkpoint is then an error.
A checkpoint is recognised in the weights folder by
hiera_large, hiera_base_plus, hiera_small or hiera_tiny in its
.pt (or .pth) file name, so renamed copies such as
my_sam2.1_hiera_large.pt are found too. The best model present is used
(large, then base_plus, small, tiny), or the one chosen with model;
the module checks that the file really holds that model. If a folder
holds several files of the chosen model, the first in alphabetical order
is used, with a warning; SAM 2.1 files (sam2.1_*) sort before SAM 2
files (sam2_*).
The region is processed in crops. A region no larger than tile_size cells is one crop, a larger region is split into tiles of tile_size cells overlapping by tile_overlap cells. With crop_layers greater than 0, each tile is additionally segmented in zoomed-in sub-crops (4 in layer 1, 16 in layer 2, ...), prompted with points_per_side divided by crop_points_downscale per layer.
Each crop is resized to 1024 x 1024 pixels, regardless of its aspect ratio, as SAM 2 does, and passed through the image encoder. The mask decoder is then prompted with a grid of points_per_side x points_per_side points, batch prompts at a time, returning three candidate masks per point. A mask is kept if its predicted IoU exceeds iou_threshold and its stability score (the IoU of the mask thresholded at plus and minus stability_offset) is at least stability_threshold. Duplicates whose bounding boxes overlap by more than nms_threshold (IoU) are suppressed, keeping the best one. Masks touching an inner tile or crop edge are discarded, since the neighbouring tile sees them whole. Holes and islands smaller than minsize cells are finally removed from each mask.
These steps and their defaults are those of the reference
SAM2AutomaticMaskGenerator, with one difference: SAM 2's generator no
longer removes small holes and islands at full resolution (its
min_mask_region_area cleans the low-resolution logits with a CUDA-only
kernel instead, skipped without CUDA). i.sam2.opencl keeps the original
SAM full-resolution cleaning, controlled by minsize. With
minsize=0, it returns the same masks as the reference generator on
identical input.
SAM 2 masks overlap: a field and the parcels inside it can both be masks. The overlap rule chooses which mask a cell belongs to: the smallest covering mask (small, finest partition, the default), the largest (large) or the one with the highest predicted IoU (score). Masks larger than maxsize cells are discarded first; this is the way to drop the single mask SAM 2 often returns for the whole background of a scene.
The resulting label raster is cleaned: every connected piece smaller than minsize cells, including small unsegmented gaps, is merged into the neighbouring piece it shares the longest border with. Boundaries are traced along cell edges into shared topological boundaries, simplified with the Douglas-Peucker algorithm using the simplify tolerance (one cell by default; 0 keeps the raster staircase), and the topology is cleaned (boundaries broken at intersections, duplicates removed, pseudo-nodes merged, slivers smaller than one cell removed). Each segment gets one centroid, placed at the cell farthest from its border.
Areas without a centroid are gaps that no mask covers. The optional raster output holds the segment categories of the cleaned label raster before simplification.
| Column | Description |
|---|---|
| cat | Segment category |
| mask | Number of the SAM 2 mask the segment comes from |
| predicted_iou | Mask quality predicted by SAM 2 |
| stability | Stability score of the mask |
| cells | Segment size in cells |
| area | Area of the simplified polygon in square map units |
A mask split into several pieces by overlapping masks yields several segments with the same mask number.
Compared with the original SAM (i.sam.opencl), SAM 2 has a hierarchical Hiera image encoder (windowed attention at four resolutions, with query pooling between them) and a feature pyramid whose two finest levels are fed to the mask decoder to sharpen mask edges. Its encoders are smaller and faster than SAM's ViT models of similar quality. SAM 2 squashes each crop to 1024 x 1024 instead of padding it, and its predicted IoU goes through a sigmoid, hence the lower default iou_threshold (0.8 instead of 0.88).
With device=auto, the first GPU is used, else the first CPU
device. When several OpenCL platforms expose the same device (for
example Mesa's Clover and rusticl drivers for AMD GPUs), the first
platform on which the kernels compile is used; platform restricts
the choice to platforms whose name contains the given text. Mesa's
rusticl only exposes a GPU when enabled in the environment, e.g.
RUSTICL_ENABLE=radeonsi. A CPU OpenCL implementation such as PoCL
gives the same result, much more slowly.
The largest work buffers are those of the first stage of the image encoder, at 256 x 256 tokens. With batch=32, the work buffers take 2330 MiB for Hiera-L and 1870 MiB for Hiera-T, in addition to the weights; a smaller batch reduces the mask decoder's share. Buffers are kept below the device's maximum allocation size.
The profile chooses how buffers are allocated and which matrix product kernels run, to suit the device:
With profile=auto the profile is detected: cpu for a CPU device, igpu for a GPU that reports unified memory, egpu for a GPU whose PCI function Linux marks as removable (when the driver reports its PCI address), dgpu otherwise. The chosen profile and the reason are printed. A profile can be forced to compare them; every profile gives the same results, only the speed differs.
precision=fp16 stores the weight matrices of the linear layers
(most of the model's memory) as IEEE half-precision numbers, read with
the core OpenCL vload_half functions, so no half-precision extension
is needed; the arithmetic stays in single precision. It halves their
memory and the bytes read by the matrix products, which matters on
devices limited by memory bandwidth (integrated GPUs, CPUs). Rounding
the weights to half precision changes the masks slightly; check the
result against precision=fp32 (the default) on a sample before
relying on it.
On a CPU (Intel Core i7-1165G7, 4 cores, PoCL), the image encoder took 46 s for Hiera-T, 116 s for Hiera-B+ and 331 s for Hiera-L per crop, and the mask decoder about 31 s per 64 prompts. A GPU is expected to be one to two orders of magnitude faster. The number of crops, and so the run time, grows with the region size divided by tile_size, and with crop_layers.
SAM 2 sees every crop at 1024 x 1024 pixels. Objects should span at least ten or so pixels at that scale: set the region resolution and tile_size accordingly. A smaller tile_size zooms in (crops are upsampled to 1024 pixels), which finds smaller objects at the cost of more crops. Square tiles avoid distorting objects: a crop much longer than wide is squashed to a square.
When the environment variable I_SAM2_OPENCL_DUMP names a directory,
the 8-bit crops, the encoder inputs, the image embeddings, the
high-resolution features and the list of masks are written there. The
script tests/reference_compare.py in the source tree compares them
with the PyTorch reference implementation.
I_SAM2_OPENCL_PROFILE=1 prints the time spent in each OpenCL kernel. Each line gives the event time and the wall-clock time (some drivers, such as rusticl on Intel iris, report wrong event times), and for the matrix products the achieved GFLOP/s, which shows whether a kernel is limited by compute or by memory bandwidth.
I_SAM2_OPENCL_DOWNLOAD_URL replaces the download location
(https://dl.fbaipublicfiles.com/segment_anything_2/092824), for a local
mirror.
Segment a true-colour composite of Landsat 7 bands 3, 2 and 1, dropping the background mask (the region has 475 x 527 cells):
g.region raster=lsat7_2002_30 -p
i.group group=lsat_rgb input=lsat7_2002_30,lsat7_2002_20,lsat7_2002_10
i.sam2.opencl group=lsat_rgb checkpoint=sam2.1_hiera_large.pt \
output=lsat_segments maxsize=50000
v.db.select lsat_segments | head
Use the fastest model for a quick look; it is downloaded into
~/models/sam/sam2 the first time:
i.sam2.opencl group=lsat_rgb model=tiny output=lsat_quick maxsize=50000
Delineate fields in a 2.5 m VNIR image with bands 1-3 grouped as RGB, keeping only segments of at least 0.1 ha (160 cells) and a label raster alongside the vector map:
i.group group=vnir_rgb input=vnir.1,vnir.2,vnir.3
g.region raster=vnir.1 -p
i.sam2.opencl group=vnir_rgb checkpoint=sam2.1_hiera_large.pt \
output=fields raster=fields minsize=160
Ravi, N., Gabeur, V., Hu, Y.-T., Hu, R., Ryali, C., Ma, T., Khedr, H., Radle, R., Rolland, C., Gustafson, L., Mintun, E., Pan, J., Alwala, K. V., Carion, N., Wu, C.-Y., Girshick, R., Dollar, P., Feichtenhofer, C., 2024. SAM 2: Segment Anything in Images and Videos. arXiv:2408.00714
Ryali, C., Hu, Y.-T., Bolya, D., Wei, C., Fan, H., Huang, P.-Y., Aggarwal, V., Chowdhury, A., Poursaeed, O., Hoffman, J., Malik, J., Li, Y., Feichtenhofer, C., 2023. Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles. International Conference on Machine Learning (ICML). arXiv:2306.00989
i.group, i.sam.opencl, i.segment, r.to.vect, v.clean, v.generalize
Yann Chemin
Available at: i.sam2.opencl source code (history)
Accessed: Sunday Nov 15 10:57:07 2026
Main index | Imagery index | Topics index | Keywords index | Graphical index | Full index
© 2003-2026 GRASS Development Team, GRASS 8.5.1dev Reference Manual