hooks: add hook for sam2

Add hook for `sam2` (Segment Anything Model 2).

_pyinstaller_hooks_contrib/stdhooks/hook-sam2.py

@@ -0,0 +1,30 @@
+# ------------------------------------------------------------------
+# Copyright (c) 2025 PyInstaller Development Team.
+#
+# This file is distributed under the terms of the GNU General Public
+# License (version 2.0 or later).
+#
+# The full license is available in LICENSE, distributed with
+# this software.
+#
+# SPDX-License-Identifier: GPL-2.0-or-later
+# ------------------------------------------------------------------
+
+# Hook for Segment Anything Model 2 (SAM 2): https://pypi.org/project/sam2
+
+from PyInstaller.utils.hooks import collect_data_files, collect_submodules
+
+# Collect config .yaml files.
+datas = collect_data_files('sam2')
+
+# Ensure that all indirectly-imported modules are collected (e.g., `sam2.modeling.backbones`).
+hiddenimports = collect_submodules('sam2')
+
+# Due to use of `torch.script`, we need to collect source .py files for `sam2`. The `sam2/__init__.py` also seems to be
+# required by `hydra`. Furthermore, the source-based introspection attempts to load the source of stdlib `enum` module.
+# The module collection mode support and run-time discovery of source .py files for modules that are collected into
+# `base_library.zip` archive was added in pyinstaller/pyinstaller#8971 (i.e., PyInstaller > 6.11.1).
+module_collection_mode = {
+    'sam2': 'pyz+py',
+    'enum': 'pyz+py',  # requires PyInstaller > 6.11.1; no-op in earlier versions
+}

news/847.new.rst

@@ -0,0 +1 @@
+Add hook for ``sam2`` (Segment Anything Model 2).

tests/test_deep_learning.py

@@ -469,3 +469,24 @@ def forward(self, x, y):
 
             assert (z == x + y).all()
     """, app_name="model_test", app_args=[str(model_file)])
+
+
+# Basic test for Segment Anything Model 2 (SAM 2)
+@importorskip('torch')
+@importorskip('sam2')
+def test_sam2(pyi_builder):
+    pyi_builder.test_source("""
+        from sam2.build_sam import build_sam2
+        from sam2.sam2_image_predictor import SAM2ImagePredictor
+
+        # No checkpoint data (= untrained model), since we would have to download it.
+        checkpoint = None
+
+        # Run on CPU to make test applicable to all OSes (default device is CUDA).
+        device = 'cpu'
+
+        model_cfg = "configs/sam2.1/sam2.1_hiera_t.yaml"
+        predictor = SAM2ImagePredictor(build_sam2(model_cfg, checkpoint, device=device))
+
+        print(predictor)
+    """)