init

app.py

@@ -0,0 +1,79 @@
+from pathlib import Path
+import torch
+from monai.bundle import ConfigParser
+import gradio as gr
+import pickle
+import torchvision.transforms as T
+import numpy as np
+import random
+
+parser = ConfigParser()
+parser.read_config(f="configs/inference.json")
+parser.read_meta(f="configs/metadata.json")
+
+inference = parser.get_parsed_content("inferer")
+# loader = parser.get_parsed_content("dataloader")
+network = parser.get_parsed_content("network_def")
+preprocess = parser.get_parsed_content("preprocessing")
+postprocess = parser.get_parsed_content("postprocessing")
+
+state_dict = torch.load("models/model.pt")
+network.load_state_dict(state_dict, strict=True)
+
+label2color = {0: (0, 0, 0),
+             1: (225, 24, 69), # RED
+             2: (135, 233, 17), # GREEN
+             3: (0, 87, 233), # BLUE
+             4: (242, 202, 25), # YELLOW
+             5: (137, 49, 239),} # PURPLE
+
+example_files = list(Path("sample_data").glob("*.png"))
+
+def visualize_instance_seg_mask(mask):
+    image = np.zeros((mask.shape[0], mask.shape[1], 3))
+    labels = np.unique(mask)
+    for i in range(image.shape[0]):
+      for j in range(image.shape[1]):
+        image[i, j, :] = label2color[mask[i, j]]
+    image = image / 255
+    return image
+
+def query_image(img, progress=gr.Progress(track_tqdm=True)):
+    data = {"image": img}
+    batch = preprocess(data)
+    
+    # with open('filename.pickle', 'rb') as handle:
+    #     pred = pickle.load(handle)
+    # batch["pred"] = pred
+    
+    network.eval()
+    with torch.no_grad():
+        pred = inference(batch['image'].unsqueeze(dim=0), network)
+        
+    batch["pred"] = pred
+    for k,v in batch["pred"].items():
+        batch["pred"][k] = v.squeeze(dim=0)
+    
+    batch = postprocess(batch)
+    
+    result = visualize_instance_seg_mask(batch["type_map"].squeeze())
+    
+    # Combine image
+    result = batch["image"].permute(1, 2, 0).cpu().numpy() * 0.5 + result * 0.5
+    
+    # Solve rotating problem
+    result = np.fliplr(result)
+    result = np.rot90(result, k=1)
+    
+    return result
+    
+demo = gr.Interface(
+    query_image, 
+    inputs=[gr.Image(type="filepath")], 
+    outputs="image",
+    title="Medical Image Classification with MONAI - Pathology Nuclei Segmentation Classification",
+    description = "Please upload an image to see segmentation capabilities of this model",
+    examples=example_files
+)
+
+demo.queue(concurrency_count=20).launch()

configs/evaluate.json

@@ -0,0 +1,38 @@
+{
+    "network_def": {
+        "_target_": "HoVerNet",
+        "mode": "@hovernet_mode",
+        "adapt_standard_resnet": true,
+        "in_channels": 3,
+        "out_classes": 5
+    },
+    "validate#handlers": [
+        {
+            "_target_": "CheckpointLoader",
+            "load_path": "$os.path.join(@bundle_root, 'models', 'model.pt')",
+            "load_dict": {
+                "model": "@network"
+            }
+        },
+        {
+            "_target_": "StatsHandler",
+            "iteration_log": false
+        },
+        {
+            "_target_": "MetricsSaver",
+            "save_dir": "@output_dir",
+            "metrics": [
+                "val_mean_dice"
+            ],
+            "metric_details": [
+                "val_mean_dice"
+            ],
+            "batch_transform": "$monai.handlers.from_engine(['image_meta_dict'])",
+            "summary_ops": "*"
+        }
+    ],
+    "evaluating": [
+        "$setattr(torch.backends.cudnn, 'benchmark', True)",
+        "$@validate#evaluator.run()"
+    ]
+}

configs/inference-Copy1.json

@@ -0,0 +1,131 @@
+{
+    "imports": [
+        "$import glob",
+        "$import os"
+    ],
+    "bundle_root": "$os.getcwd()",
+    "output_dir": ".",
+    "dataset_dir": "CoNSeP/Test/Images",
+    "num_cpus": 6,
+    "batch_size": 1,
+    "sw_batch_size": 16,
+    "hovernet_mode": "fast",
+    "patch_size": 256,
+    "out_size": 164,
+    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
+    "network_def": {
+        "_target_": "HoVerNet",
+        "mode": "@hovernet_mode",
+        "adapt_standard_resnet": true,
+        "in_channels": 3,
+        "out_classes": 5
+    },
+    "network": "$@network_def.to(@device)",
+    "preprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "LoadImaged",
+                "keys": "image",
+                "reader": "$monai.data.PILReader",
+                "converter": "$lambda x: x.convert('RGB')"
+            },
+            {
+                "_target_": "EnsureChannelFirstd",
+                "keys": "image"
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": "image",
+                "dtype": "float32"
+            },
+            {
+                "_target_": "ScaleIntensityRanged",
+                "keys": "image",
+                "a_min": 0.0,
+                "a_max": 255.0,
+                "b_min": 0.0,
+                "b_max": 1.0,
+                "clip": true
+            }
+        ]
+    },
+    "data_list": "$[{'image': image} for image in glob.glob(os.path.join(@dataset_dir, '*.png'))]",
+    "dataset": {
+        "_target_": "Dataset",
+        "data": "@data_list",
+        "transform": "@preprocessing"
+    },
+    "dataloader": {
+        "_target_": "DataLoader",
+        "dataset": "@dataset",
+        "batch_size": "@batch_size",
+        "shuffle": false,
+        "num_workers": "@num_cpus",
+        "pin_memory": true
+    },
+    "inferer": {
+        "_target_": "SlidingWindowHoVerNetInferer",
+        "roi_size": "@patch_size",
+        "sw_batch_size": "@sw_batch_size",
+        "overlap": "$1.0 - float(@out_size) / float(@patch_size)",
+        "padding_mode": "constant",
+        "cval": 0,
+        "progress": true,
+        "extra_input_padding": "$((@patch_size - @out_size) // 2,) * 4"
+    },
+    "postprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "FlattenSubKeysd",
+                "keys": "pred",
+                "sub_keys": [
+                    "horizontal_vertical",
+                    "nucleus_prediction",
+                    "type_prediction"
+                ],
+                "delete_keys": true
+            },
+            {
+                "_target_": "HoVerNetInstanceMapPostProcessingd",
+                "sobel_kernel_size": 21,
+                "marker_threshold": 0.4,
+                "marker_radius": 2
+            },
+            {
+                "_target_": "HoVerNetNuclearTypePostProcessingd"
+            },
+            {
+                "_target_": "FromMetaTensord",
+                "keys": [
+                    "image"
+                ]
+            }
+        ]
+    },
+    "handlers": [
+        {
+            "_target_": "CheckpointLoader",
+            "load_path": "$os.path.join(@bundle_root, 'models', 'model.pt')",
+            "map_location": "@device",
+            "load_dict": {
+                "model": "@network"
+            }
+        }
+    ],
+    "evaluator": {
+        "_target_": "SupervisedEvaluator",
+        "device": "@device",
+        "val_data_loader": "@dataloader",
+        "val_handlers": "@handlers",
+        "network": "@network",
+        "postprocessing": "@postprocessing",
+        "inferer": "@inferer",
+        "amp": true
+    },
+    "evaluating": [
+        "$setattr(torch.backends.cudnn, 'benchmark', True)",
+        "$@evaluator.run()"
+    ]
+}

configs/inference.json

@@ -0,0 +1,131 @@
+{
+    "imports": [
+        "$import glob",
+        "$import os"
+    ],
+    "bundle_root": "$os.getcwd()",
+    "output_dir": ".",
+    "dataset_dir": "CoNSeP/Test/Images",
+    "num_cpus": 2,
+    "batch_size": 1,
+    "sw_batch_size": 16,
+    "hovernet_mode": "fast",
+    "patch_size": 256,
+    "out_size": 164,
+    "device": "cpu",
+    "network_def": {
+        "_target_": "HoVerNet",
+        "mode": "@hovernet_mode",
+        "adapt_standard_resnet": true,
+        "in_channels": 3,
+        "out_classes": 5
+    },
+    "network": "$@network_def.to(@device)",
+    "preprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "LoadImaged",
+                "keys": "image",
+                "reader": "$monai.data.PILReader",
+                "converter": "$lambda x: x.convert('RGB')"
+            },
+            {
+                "_target_": "EnsureChannelFirstd",
+                "keys": "image"
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": "image",
+                "dtype": "float32"
+            },
+            {
+                "_target_": "ScaleIntensityRanged",
+                "keys": "image",
+                "a_min": 0.0,
+                "a_max": 255.0,
+                "b_min": 0.0,
+                "b_max": 1.0,
+                "clip": true
+            }
+        ]
+    },
+    "data_list": "$[{'image': image} for image in glob.glob(os.path.join(@dataset_dir, '*.png'))]",
+    "dataset": {
+        "_target_": "Dataset",
+        "data": "@data_list",
+        "transform": "@preprocessing"
+    },
+    "dataloader": {
+        "_target_": "DataLoader",
+        "dataset": "@dataset",
+        "batch_size": "@batch_size",
+        "shuffle": false,
+        "num_workers": "@num_cpus",
+        "pin_memory": true
+    },
+    "inferer": {
+        "_target_": "SlidingWindowHoVerNetInferer",
+        "roi_size": "@patch_size",
+        "sw_batch_size": "@sw_batch_size",
+        "overlap": "$1.0 - float(@out_size) / float(@patch_size)",
+        "padding_mode": "constant",
+        "cval": 0,
+        "progress": true,
+        "extra_input_padding": "$((@patch_size - @out_size) // 2,) * 4"
+    },
+    "postprocessing": {
+        "_target_": "Compose",
+        "transforms": [
+            {
+                "_target_": "FlattenSubKeysd",
+                "keys": "pred",
+                "sub_keys": [
+                    "horizontal_vertical",
+                    "nucleus_prediction",
+                    "type_prediction"
+                ],
+                "delete_keys": true
+            },
+            {
+                "_target_": "HoVerNetInstanceMapPostProcessingd",
+                "sobel_kernel_size": 21,
+                "marker_threshold": 0.4,
+                "marker_radius": 2
+            },
+            {
+                "_target_": "HoVerNetNuclearTypePostProcessingd"
+            },
+            {
+                "_target_": "FromMetaTensord",
+                "keys": [
+                    "image"
+                ]
+            }
+        ]
+    },
+    "handlers": [
+        {
+            "_target_": "CheckpointLoader",
+            "load_path": "$os.path.join(@bundle_root, 'models', 'model.pt')",
+            "map_location": "@device",
+            "load_dict": {
+                "model": "@network"
+            }
+        }
+    ],
+    "evaluator": {
+        "_target_": "SupervisedEvaluator",
+        "device": "@device",
+        "val_data_loader": "@dataloader",
+        "val_handlers": "@handlers",
+        "network": "@network",
+        "postprocessing": "@postprocessing",
+        "inferer": "@inferer",
+        "amp": true
+    },
+    "evaluating": [
+        "$setattr(torch.backends.cudnn, 'benchmark', True)",
+        "$@evaluator.run()"
+    ]
+}

configs/logging.conf

@@ -0,0 +1,21 @@
+[loggers]
+keys=root
+
+[handlers]
+keys=consoleHandler
+
+[formatters]
+keys=fullFormatter
+
+[logger_root]
+level=INFO
+handlers=consoleHandler
+
+[handler_consoleHandler]
+class=StreamHandler
+level=INFO
+formatter=fullFormatter
+args=(sys.stdout,)
+
+[formatter_fullFormatter]
+format=%(asctime)s - %(name)s - %(levelname)s - %(message)s

configs/metadata.json

@@ -0,0 +1,119 @@
+{
+    "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_hovernet_20221124.json",
+    "version": "0.1.3",
+    "changelog": {
+        "0.1.3": "add name tag",
+        "0.1.2": "update the workflow figure",
+        "0.1.1": "update to use monai 1.1.0",
+        "0.1.0": "complete the model package"
+    },
+    "monai_version": "1.1.0",
+    "pytorch_version": "1.13.0",
+    "numpy_version": "1.22.2",
+    "optional_packages_version": {
+        "scikit-image": "0.19.3",
+        "scipy": "1.8.1",
+        "tqdm": "4.64.1",
+        "pillow": "9.0.1"
+    },
+    "name": "Nuclear segmentation and classification",
+    "task": "Nuclear segmentation and classification",
+    "description": "A simultaneous segmentation and classification of nuclei within multitissue histology images based on CoNSeP data",
+    "authors": "MONAI team",
+    "copyright": "Copyright (c) MONAI Consortium",
+    "data_source": "https://warwick.ac.uk/fac/cross_fac/tia/data/hovernet/",
+    "data_type": "numpy",
+    "image_classes": "RGB image with intensity between 0 and 255",
+    "label_classes": "a dictionary contains binary nuclear segmentation, hover map and pixel-level classification",
+    "pred_classes": "a dictionary contains scalar probability for binary nuclear segmentation, hover map and pixel-level classification",
+    "eval_metrics": {
+        "Binary Dice": 0.8293,
+        "PQ": 0.4936,
+        "F1d": 0.748
+    },
+    "intended_use": "This is an example, not to be used for diagnostic purposes",
+    "references": [
+        "Simon Graham. 'HoVer-Net: Simultaneous Segmentation and Classification of Nuclei in Multi-Tissue Histology Images.' Medical Image Analysis, 2019. https://arxiv.org/abs/1812.06499"
+    ],
+    "network_data_format": {
+        "inputs": {
+            "image": {
+                "type": "image",
+                "format": "magnitude",
+                "num_channels": 3,
+                "spatial_shape": [
+                    "256",
+                    "256"
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    255
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "image"
+                }
+            }
+        },
+        "outputs": {
+            "nucleus_prediction": {
+                "type": "probability",
+                "format": "segmentation",
+                "num_channels": 3,
+                "spatial_shape": [
+                    "164",
+                    "164"
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    1
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "background",
+                    "1": "nuclei"
+                }
+            },
+            "horizontal_vertical": {
+                "type": "probability",
+                "format": "regression",
+                "num_channels": 2,
+                "spatial_shape": [
+                    "164",
+                    "164"
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    1
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "horizontal distances map",
+                    "1": "vertical distances map"
+                }
+            },
+            "type_prediction": {
+                "type": "probability",
+                "format": "classification",
+                "num_channels": 2,
+                "spatial_shape": [
+                    "164",
+                    "164"
+                ],
+                "dtype": "float32",
+                "value_range": [
+                    0,
+                    1
+                ],
+                "is_patch_data": true,
+                "channel_def": {
+                    "0": "background",
+                    "1": "type of nucleus for each pixel"
+                }
+            }
+        }
+    }
+}

configs/multi_gpu_train.json

@@ -0,0 +1,36 @@
+{
+    "device": "$torch.device(f'cuda:{dist.get_rank()}')",
+    "network": {
+        "_target_": "torch.nn.parallel.DistributedDataParallel",
+        "module": "$@network_def.to(@device)",
+        "device_ids": [
+            "@device"
+        ]
+    },
+    "train#sampler": {
+        "_target_": "DistributedSampler",
+        "dataset": "@train#dataset",
+        "even_divisible": true,
+        "shuffle": true
+    },
+    "train#dataloader#sampler": "@train#sampler",
+    "train#dataloader#shuffle": false,
+    "train#trainer#train_handlers": "$@train#train_handlers[: -2 if dist.get_rank() > 0 else None]",
+    "validate#sampler": {
+        "_target_": "DistributedSampler",
+        "dataset": "@validate#dataset",
+        "even_divisible": false,
+        "shuffle": false
+    },
+    "validate#dataloader#sampler": "@validate#sampler",
+    "validate#evaluator#val_handlers": "$None if dist.get_rank() > 0 else @validate#handlers",
+    "training": [
+        "$import torch.distributed as dist",
+        "$dist.init_process_group(backend='nccl')",
+        "$torch.cuda.set_device(@device)",
+        "$monai.utils.set_determinism(seed=321)",
+        "$setattr(torch.backends.cudnn, 'benchmark', True)",
+        "$@train#trainer.run()",
+        "$dist.destroy_process_group()"
+    ]
+}

configs/train.json

@@ -0,0 +1,525 @@
+{
+    "imports": [
+        "$import glob",
+        "$import os",
+        "$import skimage"
+    ],
+    "bundle_root": "$os.getcwd()",
+    "ckpt_dir_stage0": "$os.path.join(@bundle_root, 'models', 'stage0')",
+    "ckpt_dir_stage1": "$os.path.join(@bundle_root, 'models')",
+    "ckpt_path_stage0": "$os.path.join(@ckpt_dir_stage0, 'model.pt')",
+    "output_dir": "$os.path.join(@bundle_root, 'eval')",
+    "dataset_dir": "/workspace/Data/Pathology/CoNSeP/Prepared/",
+    "train_images": "$list(sorted(glob.glob(@dataset_dir + '/Train/*image.npy')))",
+    "val_images": "$list(sorted(glob.glob(@dataset_dir + '/Test/*image.npy')))",
+    "train_inst_map": "$list(sorted(glob.glob(@dataset_dir + '/Train/*inst_map.npy')))",
+    "val_inst_map": "$list(sorted(glob.glob(@dataset_dir + '/Test/*inst_map.npy')))",
+    "train_type_map": "$list(sorted(glob.glob(@dataset_dir + '/Train/*type_map.npy')))",
+    "val_type_map": "$list(sorted(glob.glob(@dataset_dir + '/Test/*type_map.npy')))",
+    "device": "$torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')",
+    "stage": 0,
+    "epochs": 50,
+    "val_interval": 1,
+    "learning_rate": 0.0001,
+    "amp": true,
+    "hovernet_mode": "fast",
+    "patch_size": 256,
+    "out_size": 164,
+    "ckpt_dir": "$@ckpt_dir_stage0 if @stage == 0 else @ckpt_dir_stage1",
+    "network_def": {
+        "_target_": "HoVerNet",
+        "mode": "@hovernet_mode",
+        "in_channels": 3,
+        "out_classes": 5,
+        "adapt_standard_resnet": true,
+        "pretrained_url": "$None",
+        "freeze_encoder": true
+    },
+    "network": "$@network_def.to(@device)",
+    "loss": {
+        "_target_": "HoVerNetLoss",
+        "lambda_hv_mse": 1.0
+    },
+    "optimizer": {
+        "_target_": "torch.optim.Adam",
+        "params": "$filter(lambda p: p.requires_grad, @network.parameters())",
+        "lr": "@learning_rate",
+        "weight_decay": 1e-05
+    },
+    "lr_scheduler": {
+        "_target_": "torch.optim.lr_scheduler.StepLR",
+        "optimizer": "@optimizer",
+        "step_size": 25
+    },
+    "train": {
+        "preprocessing_transforms": [
+            {
+                "_target_": "LoadImaged",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ]
+            },
+            {
+                "_target_": "EnsureChannelFirstd",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "channel_dim": -1
+            },
+            {
+                "_target_": "Lambdad",
+                "keys": "label_inst",
+                "func": "$lambda x: skimage.measure.label(x)"
+            },
+            {
+                "_target_": "RandAffined",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "prob": 1.0,
+                "rotate_range": [
+                    "$np.pi"
+                ],
+                "scale_range": [
+                    [
+                        -0.2,
+                        0.2
+                    ],
+                    [
+                        -0.2,
+                        0.2
+                    ]
+                ],
+                "shear_range": [
+                    [
+                        -0.05,
+                        0.05
+                    ],
+                    [
+                        -0.05,
+                        0.05
+                    ]
+                ],
+                "translate_range": [
+                    [
+                        -6,
+                        6
+                    ],
+                    [
+                        -6,
+                        6
+                    ]
+                ],
+                "padding_mode": "zeros",
+                "mode": "nearest"
+            },
+            {
+                "_target_": "CenterSpatialCropd",
+                "keys": [
+                    "image"
+                ],
+                "roi_size": [
+                    "@patch_size",
+                    "@patch_size"
+                ]
+            },
+            {
+                "_target_": "RandFlipd",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "prob": 0.5,
+                "spatial_axis": 0
+            },
+            {
+                "_target_": "RandFlipd",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "prob": 0.5,
+                "spatial_axis": 1
+            },
+            {
+                "_target_": "OneOf",
+                "transforms": [
+                    {
+                        "_target_": "RandGaussianSmoothd",
+                        "keys": [
+                            "image"
+                        ],
+                        "sigma_x": [
+                            0.1,
+                            1.1
+                        ],
+                        "sigma_y": [
+                            0.1,
+                            1.1
+                        ],
+                        "prob": 1.0
+                    },
+                    {
+                        "_target_": "MedianSmoothd",
+                        "keys": [
+                            "image"
+                        ],
+                        "radius": 1
+                    },
+                    {
+                        "_target_": "RandGaussianNoised",
+                        "keys": [
+                            "image"
+                        ],
+                        "std": 0.05,
+                        "prob": 1.0
+                    }
+                ]
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": "image",
+                "dtype": "$np.uint8"
+            },
+            {
+                "_target_": "TorchVisiond",
+                "keys": "image",
+                "name": "ColorJitter",
+                "brightness": [
+                    0.9,
+                    1.0
+                ],
+                "contrast": [
+                    0.95,
+                    1.1
+                ],
+                "saturation": [
+                    0.8,
+                    1.2
+                ],
+                "hue": [
+                    -0.04,
+                    0.04
+                ]
+            },
+            {
+                "_target_": "AsDiscreted",
+                "keys": "label_type",
+                "to_onehot": 5
+            },
+            {
+                "_target_": "ScaleIntensityRanged",
+                "keys": "image",
+                "a_min": 0.0,
+                "a_max": 255.0,
+                "b_min": 0.0,
+                "b_max": 1.0,
+                "clip": true
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": "label_inst",
+                "dtype": "$torch.int"
+            },
+            {
+                "_target_": "ComputeHoVerMapsd",
+                "keys": "label_inst"
+            },
+            {
+                "_target_": "Lambdad",
+                "keys": "label_inst",
+                "func": "$lambda x: x > 0",
+                "overwrite": "label"
+            },
+            {
+                "_target_": "CenterSpatialCropd",
+                "keys": [
+                    "label",
+                    "hover_label_inst",
+                    "label_inst",
+                    "label_type"
+                ],
+                "roi_size": [
+                    "@out_size",
+                    "@out_size"
+                ]
+            },
+            {
+                "_target_": "AsDiscreted",
+                "keys": "label",
+                "to_onehot": 2
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "dtype": "$torch.float32"
+            }
+        ],
+        "preprocessing": {
+            "_target_": "Compose",
+            "transforms": "$@train#preprocessing_transforms"
+        },
+        "dataset": {
+            "_target_": "Dataset",
+            "data": "$[{'image': i, 'label_inst': j, 'label_type': k} for i, j, k in zip(@train_images, @train_inst_map, @train_type_map)]",
+            "transform": "@train#preprocessing"
+        },
+        "dataloader": {
+            "_target_": "DataLoader",
+            "dataset": "@train#dataset",
+            "batch_size": 16,
+            "shuffle": true,
+            "num_workers": 4
+        },
+        "inferer": {
+            "_target_": "SimpleInferer"
+        },
+        "postprocessing_np": {
+            "_target_": "Compose",
+            "transforms": [
+                {
+                    "_target_": "Activationsd",
+                    "keys": "nucleus_prediction",
+                    "softmax": true
+                },
+                {
+                    "_target_": "AsDiscreted",
+                    "keys": "nucleus_prediction",
+                    "argmax": true
+                }
+            ]
+        },
+        "postprocessing": {
+            "_target_": "Lambdad",
+            "keys": "pred",
+            "func": "$@train#postprocessing_np"
+        },
+        "handlers": [
+            {
+                "_target_": "LrScheduleHandler",
+                "lr_scheduler": "@lr_scheduler",
+                "print_lr": true
+            },
+            {
+                "_target_": "ValidationHandler",
+                "validator": "@validate#evaluator",
+                "epoch_level": true,
+                "interval": "@val_interval"
+            },
+            {
+                "_target_": "CheckpointSaver",
+                "save_dir": "@ckpt_dir",
+                "save_dict": {
+                    "model": "@network"
+                },
+                "save_interval": 10,
+                "epoch_level": true,
+                "save_final": true,
+                "final_filename": "model.pt"
+            },
+            {
+                "_target_": "StatsHandler",
+                "tag_name": "train_loss",
+                "output_transform": "$monai.handlers.from_engine(['loss'], first=True)"
+            },
+            {
+                "_target_": "TensorBoardStatsHandler",
+                "log_dir": "@output_dir",
+                "tag_name": "train_loss",
+                "output_transform": "$monai.handlers.from_engine(['loss'], first=True)"
+            }
+        ],
+        "extra_handlers": [
+            {
+                "_target_": "CheckpointLoader",
+                "load_path": "$os.path.join(@ckpt_dir_stage0, 'model.pt')",
+                "load_dict": {
+                    "model": "@network"
+                }
+            }
+        ],
+        "train_handlers": "$@train#extra_handlers + @train#handlers if @stage==1 else @train#handlers",
+        "key_metric": {
+            "train_mean_dice": {
+                "_target_": "MeanDice",
+                "include_background": false,
+                "output_transform": "$monai.apps.pathology.handlers.utils.from_engine_hovernet(keys=['pred', 'label'], nested_key='nucleus_prediction')"
+            }
+        },
+        "trainer": {
+            "_target_": "SupervisedTrainer",
+            "max_epochs": "@epochs",
+            "device": "@device",
+            "train_data_loader": "@train#dataloader",
+            "prepare_batch": "$monai.apps.pathology.engines.utils.PrepareBatchHoVerNet(extra_keys=['label_type', 'hover_label_inst'])",
+            "network": "@network",
+            "loss_function": "@loss",
+            "optimizer": "@optimizer",
+            "inferer": "@train#inferer",
+            "postprocessing": "@train#postprocessing",
+            "key_train_metric": "@train#key_metric",
+            "train_handlers": "@train#train_handlers",
+            "amp": "@amp"
+        }
+    },
+    "validate": {
+        "preprocessing_transforms": [
+            {
+                "_target_": "LoadImaged",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ]
+            },
+            {
+                "_target_": "EnsureChannelFirstd",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "channel_dim": -1
+            },
+            {
+                "_target_": "Lambdad",
+                "keys": "label_inst",
+                "func": "$lambda x: skimage.measure.label(x)"
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": [
+                    "image",
+                    "label_inst"
+                ],
+                "dtype": "$torch.int"
+            },
+            {
+                "_target_": "CenterSpatialCropd",
+                "keys": [
+                    "image"
+                ],
+                "roi_size": [
+                    "@patch_size",
+                    "@patch_size"
+                ]
+            },
+            {
+                "_target_": "ScaleIntensityRanged",
+                "keys": "image",
+                "a_min": 0.0,
+                "a_max": 255.0,
+                "b_min": 0.0,
+                "b_max": 1.0,
+                "clip": true
+            },
+            {
+                "_target_": "ComputeHoVerMapsd",
+                "keys": "label_inst"
+            },
+            {
+                "_target_": "Lambdad",
+                "keys": "label_inst",
+                "func": "$lambda x: x > 0",
+                "overwrite": "label"
+            },
+            {
+                "_target_": "CenterSpatialCropd",
+                "keys": [
+                    "label",
+                    "hover_label_inst",
+                    "label_inst",
+                    "label_type"
+                ],
+                "roi_size": [
+                    "@out_size",
+                    "@out_size"
+                ]
+            },
+            {
+                "_target_": "CastToTyped",
+                "keys": [
+                    "image",
+                    "label_inst",
+                    "label_type"
+                ],
+                "dtype": "$torch.float32"
+            }
+        ],
+        "preprocessing": {
+            "_target_": "Compose",
+            "transforms": "$@validate#preprocessing_transforms"
+        },
+        "dataset": {
+            "_target_": "Dataset",
+            "data": "$[{'image': i, 'label_inst': j, 'label_type': k} for i, j, k in zip(@val_images, @val_inst_map, @val_type_map)]",
+            "transform": "@validate#preprocessing"
+        },
+        "dataloader": {
+            "_target_": "DataLoader",
+            "dataset": "@validate#dataset",
+            "batch_size": 16,
+            "shuffle": false,
+            "num_workers": 4
+        },
+        "inferer": {
+            "_target_": "SimpleInferer"
+        },
+        "postprocessing": "$@train#postprocessing",
+        "handlers": [
+            {
+                "_target_": "StatsHandler",
+                "iteration_log": false
+            },
+            {
+                "_target_": "TensorBoardStatsHandler",
+                "log_dir": "@output_dir",
+                "iteration_log": false
+            },
+            {
+                "_target_": "CheckpointSaver",
+                "save_dir": "@ckpt_dir",
+                "save_dict": {
+                    "model": "@network"
+                },
+                "save_key_metric": true
+            }
+        ],
+        "key_metric": {
+            "val_mean_dice": {
+                "_target_": "MeanDice",
+                "include_background": false,
+                "output_transform": "$monai.apps.pathology.handlers.utils.from_engine_hovernet(keys=['pred', 'label'], nested_key='nucleus_prediction')"
+            }
+        },
+        "evaluator": {
+            "_target_": "SupervisedEvaluator",
+            "device": "@device",
+            "val_data_loader": "@validate#dataloader",
+            "prepare_batch": "$monai.apps.pathology.engines.utils.PrepareBatchHoVerNet(extra_keys=['label_type', 'hover_label_inst'])",
+            "network": "@network",
+            "inferer": "@validate#inferer",
+            "postprocessing": "@validate#postprocessing",
+            "key_val_metric": "@validate#key_metric",
+            "val_handlers": "@validate#handlers",
+            "amp": "@amp"
+        }
+    },
+    "training": [
+        "$monai.utils.set_determinism(seed=321)",
+        "$setattr(torch.backends.cudnn, 'benchmark', True)",
+        "$@train#trainer.run()"
+    ]
+}

requirements.txt

@@ -0,0 +1,2 @@
+monai==1.1.0
+scikit-image==0.20.0