Merge pull request #2 from ThuanNaN/create_dataset

Add MVTEC dataset

.gitignore

@@ -163,7 +163,6 @@ cython_debug/
 wandb/
 jobs/logs/
 *.out
-*ipynb
 .history/
 *.json
 *.sh
@@ -182,3 +181,7 @@ sbatch_generate*
 eval_data/
 dataset/Evaluation.md
 jupyter_notebook.slurm
+MVTEC/
+MVTEC_det/
+*.pth
+log.*

eval_configs/minigptv2_eval_mvtec.yaml

@@ -0,0 +1,25 @@
+model:
+  arch: minigpt_v2
+  model_type: pretrain
+  max_txt_len: 500
+  end_sym: "</s>"
+  low_resource: True
+  prompt_template: '[INST] {} [/INST]'
+  # ckpt: "./minigpt4/mvtec_outputs/20250112105/checkpoint_9.pth"
+  ckpt: "./ckpt/checkpoint_stage3.pth"
+  lora_r: 64
+  lora_alpha: 16
+
+
+datasets:
+  cc_sbu_align:
+    vis_processor:
+      train:
+        name: "blip2_image_eval"
+        image_size: 448
+    text_processor:
+      train:
+        name: "blip_caption"
+
+run:
+  task: image_text_pretrain

minigpt4/configs/datasets/mvtec/default.yaml

@@ -0,0 +1,6 @@
+datasets:
+  mvtec_ad:
+    data_type: images
+    build_info:
+      image_path: ./MVTEC_det/images
+      ann_path: ./MVTEC_det/mvtech_ad_data_for_regression.json

minigpt4/configs/models/minigpt_v2.yaml

@@ -11,7 +11,7 @@ model:
   # generation configs
   prompt: ""
 
-  llama_model: "please set this value to the path of llama2-chat-7b"
+  llama_model: "meta-llama/Llama-2-7b-chat-hf"
   lora_r: 64
   lora_alpha: 16
 

minigpt4/datasets/builders/image_text_pair_builder.py

@@ -18,7 +18,7 @@ from minigpt4.datasets.datasets.aok_vqa_datasets import AOKVQADataset
 from minigpt4.datasets.datasets.coco_vqa_datasets import COCOVQADataset
 from minigpt4.datasets.datasets.ocrvqa_dataset import OCRVQADataset
 from minigpt4.datasets.datasets.coco_caption import COCOCapDataset
-
+from minigpt4.datasets.datasets.mvtec_dataset import MVTecDataset
 
 @registry.register_builder("multitask_conversation")
 class MultitaskConversationBuilder(BaseDatasetBuilder):
@@ -394,6 +394,29 @@ class CaptionToPhraseBuilder(BaseDatasetBuilder):
         return datasets
 
 
+@registry.register_builder("mvtec_ad")
+class MVTECADBuilder(BaseDatasetBuilder):
+    train_dataset_cls = MVTecDataset
+    DATASET_CONFIG_DICT = {
+        "default": "configs/datasets/mvtec/default.yaml",
+    }
+    def build_datasets(self):
+        logging.info("Building datasets...")
+        self.build_processors()
+        build_info = self.config.build_info
+        datasets = dict()
+
+        # create datasets
+        dataset_cls = self.train_dataset_cls
+        datasets['train'] = dataset_cls(
+            vis_processor=self.vis_processors["train"],
+            text_processor=self.text_processors["train"],
+            ann_path=build_info.ann_path,
+            vis_root=build_info.image_path,
+        )
+
+        return datasets
+
 
 
 class DocumentVQABuilder(BaseDatasetBuilder):

minigpt4/datasets/datasets/mvtec_dataset.py

@@ -0,0 +1,51 @@
+import os
+import json
+import random
+from PIL import Image
+from torch.utils.data import Dataset
+
+
+class MVTecDataset(Dataset):
+    def __init__(self, vis_processor, text_processor, vis_root, ann_path):
+        """
+        vis_root (string): Root directory of images (e.g. coco/images/)
+        ann_root (string): directory to store the annotation file
+        """
+        self.vis_root = vis_root
+
+        self.vis_processor = vis_processor
+        self.text_processor = text_processor
+
+        self.instruction_pool = [
+            '[detection] {}',
+        ]
+
+        with open(ann_path, 'r') as f:
+            self.ann = json.load(f)
+
+    def __len__(self):
+        return len(self.ann)
+
+    def __getitem__(self, index):
+        info = self.ann[index]
+        gt_bbox = info["bbox"]
+
+        image_path = os.path.join(self.vis_root, info['image_path'])
+        image = Image.open(image_path).convert("RGB")
+        image = self.vis_processor(image)
+
+        input = "detect defect or non-defect and return the bounding box"
+        
+        ans_cls = "defect" if info["is_broken"] == True else "non-defect"
+        answer = f"{ans_cls}<{gt_bbox[0]}><{gt_bbox[1]}><{gt_bbox[2]}><{gt_bbox[3]}>"
+
+        instruction = random.choice(self.instruction_pool).format(input)
+        instruction = "<Img><ImageHere></Img> {} ".format(instruction)
+
+        return {
+            "image": image,
+            "instruction_input": instruction,
+            "answer": answer,
+            "image_id": info['image_path'],
+        }
+    

train_configs/minigptv2_finetune_mvtec.yaml

@@ -0,0 +1,55 @@
+model:
+  arch: minigpt_v2
+  model_type: pretrain
+  max_txt_len: 1024
+  image_size: 448
+  end_sym: "</s>"
+  llama_model: "meta-llama/Llama-2-7b-chat-hf"
+  ckpt: "./ckpt/checkpoint_stage3.pth"
+  use_grad_checkpoint: True
+  chat_template: True
+  lora_r: 64
+  lora_alpha: 16
+
+datasets:
+  mvtec_ad: 
+    batch_size: 2
+    vis_processor:
+      train:
+        name: "blip2_image_train"
+        image_size: 448
+    text_processor:
+      train:
+        name: "blip_caption"
+    sample_ratio: 100
+
+run:
+  task: image_text_pretrain
+  # optimizer
+  lr_sched: "linear_warmup_cosine_lr"
+  init_lr: 1e-5
+  min_lr: 1e-6
+  warmup_lr: 1e-6
+
+  weight_decay: 0.05
+  max_epoch: 10
+  num_workers: 6
+  warmup_steps: 1000
+  iters_per_epoch: 1000
+
+  seed: 42
+  output_dir: "mvtec_outputs"
+
+  amp: True
+  resume_ckpt_path: null
+
+  evaluate: False 
+  train_splits: ["train"]
+
+  device: "cuda"
+  world_size: 1
+  dist_url: "env://"
+  distributed: True
+
+  wandb_log: True
+  job_name: minigptv2_finetune