Add TensorRT Support to master branch (from #14)

.gitignore

@@ -0,0 +1,160 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+.idea/

README.md

@@ -17,6 +17,8 @@ estimation, loads the pre-trained weights, and make human predictions on a singl
 - Support for multi-GPU inference.
 - Multi-person support by design (HigherHRNet is a bottom-up approach).
 - A reference code that runs a live demo reading frames from a webcam or a video file.
+- **NEW** Support for TensorRT (thanks to [@gpastal24](https://github.com/gpastal24), see [#14](https://github.com/stefanopini/simple-HigherHRNet/pull/14) and [#15](https://github.com/stefanopini/simple-HigherHRNet/pull/15)).
+- **NEW** A [Jupyter Notebook](https://github.com/stefanopini/simple-HigherHRNet/blob/master/colab_notebook.ipynb) compatible with Google Colab showcasing how to use this repository.
 
 This repository is built along the lines of the repository
 [*simple-HRNet*](https://github.com/stefanopini/simple-HRNet).  
@@ -110,6 +112,24 @@ For help:
 python scripts/extract-keypoints.py --help
 ```
 
+### Converting the model to TensorRT:
+
+Warning: require the installation of TensorRT (see Nvidia website) and onnx.
+On some platforms, they can be installed with
+```
+pip install tensorrt onnx
+```
+
+Converting in FP16:
+```
+python scripts/export-tensorrt-model.py --device 0 --half
+```
+
+For help:
+```
+python scripts/export-tensorrt-model.py --help
+```
+
 ### Installation instructions
 
 - Clone the repository  
@@ -143,3 +163,5 @@ python scripts/extract-keypoints.py --help
 - [ ] Optimize the post-processing steps
 - [ ] Add COCO dataset and evaluation
 - [ ] Add Train/Test scripts
+- [x] Add TensorRT support
+- [x] Add notebook compatible with Colab

SimpleHigherHRNet.py

@@ -7,7 +7,8 @@
 
 from models.higherhrnet import HigherHRNet
 from misc.HeatmapParser import HeatmapParser
-from misc.utils import get_multi_scale_size, resize_align_multi_scale, get_multi_stage_outputs, aggregate_results, get_final_preds, bbox_iou
+from misc.utils import (get_multi_scale_size, resize_align_multi_scale, get_multi_stage_outputs, aggregate_results,
+                        get_final_preds)
 
 
 class SimpleHigherHRNet:
@@ -30,7 +31,8 @@ def __init__(self,
                  filter_redundant_poses=True,
                  max_nof_people=30,
                  max_batch_size=32,
-                 device=torch.device("cpu")):
+                 device=torch.device("cpu"),
+                 enable_tensorrt=False):
         """
         Initializes a new SimpleHigherHRNet object.
         HigherHRNet is initialized on the torch.device("device") and
@@ -60,6 +62,9 @@ def __init__(self,
                 Default: 16
             device (:class:`torch.device` or str): the higherhrnet (and yolo) inference will be run on this device.
                 Default: torch.device("cpu")
+            enable_tensorrt (bool): Enables tensorrt inference for HigherHRnet.
+                If enabled, a `.engine` file is expected as `checkpoint_path`.
+                Default: False
         """
 
         self.c = c
@@ -74,6 +79,7 @@ def __init__(self,
         self.max_nof_people = max_nof_people
         self.max_batch_size = max_batch_size
         self.device = device
+        self.enable_tensorrt = enable_tensorrt
 
         # assert nof_joints in (14, 15, 17)
         if self.nof_joints == 14:
@@ -90,33 +96,39 @@ def __init__(self,
         else:
             raise ValueError('Wrong model name.')
 
-        checkpoint = torch.load(checkpoint_path, map_location=self.device)
-        if 'model' in checkpoint:
-            checkpoint = checkpoint['model']
-        # fix issue with official high-resolution weights
-        checkpoint = OrderedDict([(k[2:] if k[:2] == '1.' else k, v) for k, v in checkpoint.items()])
-        self.model.load_state_dict(checkpoint)
+        if not self.enable_tensorrt:
+            checkpoint = torch.load(checkpoint_path, map_location=self.device)
+            if 'model' in checkpoint:
+                checkpoint = checkpoint['model']
+            # fix issue with official high-resolution weights
+            checkpoint = OrderedDict([(k[2:] if k[:2] == '1.' else k, v) for k, v in checkpoint.items()])
+            self.model.load_state_dict(checkpoint)
+            if 'cuda' in str(self.device):
+                print("device: 'cuda' - ", end="")
+
+                if 'cuda' == str(self.device):
+                    # if device is set to 'cuda', all available GPUs will be used
+                    print("%d GPU(s) will be used" % torch.cuda.device_count())
+                    device_ids = None
+                else:
+                    # if device is set to 'cuda:IDS', only that/those device(s) will be used
+                    print("GPU(s) '%s' will be used" % str(self.device))
+                    device_ids = [int(x) for x in str(self.device)[5:].split(',')]
 
-        if 'cuda' in str(self.device):
-            print("device: 'cuda' - ", end="")
+                self.model = torch.nn.DataParallel(self.model, device_ids=device_ids)
 
-            if 'cuda' == str(self.device):
-                # if device is set to 'cuda', all available GPUs will be used
-                print("%d GPU(s) will be used" % torch.cuda.device_count())
-                device_ids = None
+            elif 'cpu' == str(self.device):
+                print("device: 'cpu'")
             else:
-                # if device is set to 'cuda:IDS', only that/those device(s) will be used
-                print("GPU(s) '%s' will be used" % str(self.device))
-                device_ids = [int(x) for x in str(self.device)[5:].split(',')]
+                raise ValueError('Wrong device name.')
 
-            self.model = torch.nn.DataParallel(self.model, device_ids=device_ids)
-        elif 'cpu' == str(self.device):
-            print("device: 'cpu'")
+            self.model = self.model.to(device)
+            self.model.eval()
         else:
-            raise ValueError('Wrong device name.')
-
-        self.model = self.model.to(device)
-        self.model.eval()
+            if device.type == 'cpu':
+                raise ValueError('TensorRT does not support cpu device.')
+            from misc.tensorrt_utils import TRTModule_HigherHRNet
+            self.model = TRTModule_HigherHRNet(path=checkpoint_path, device=self.device)
 
         self.output_parser = HeatmapParser(num_joints=self.nof_joints,
                                            joint_set=self.joint_set,