掌握这个开源工具，更快速地构建计算机视觉模型

from flash.image import ImageClassificationDataimport fiftyone as fotrain_dataset = fo.Dataset.from_dir("/path/to/train",fo.types.ImageClassificationDirectoryTree,label_field="ground_truth",)val_dataset = fo.Dataset.from_dir("/path/to/val",fo.types.ImageClassificationDirectoryTree,label_field="ground_truth",)datamodule = ImageClassificationData.from_fiftyone(train_dataset=train_dataset,val_dataset=val_dataset,label_field="ground_truth",
)

import flashfrom flash.core.data.utils import download_datafrom flash.image import ImageClassificationData, ImageClassifier# 1. Download the datadownload_data("https://pl-flash-data.s3.amazonaws.com/hymenoptera_data.zip", 'data/')# 2. Load the datadatamodule = ImageClassificationData.from_folders(train_folder="data/hymenoptera_data/train/",val_folder="data/hymenoptera_data/val/",test_folder="data/hymenoptera_data/test/",)# 3. Build the modelmodel = ImageClassifier(num_classes=datamodule.num_classes, backbone="resnet18")# 4. Create the trainer. Run once on datatrainer = flash.Trainer(max_epochs=1)# 5. Finetune the modeltrainer.finetune(model, datamodule=datamodule, strategy="freeze")# 6. Save it!trainer.save_checkpoint("image_classification_model.pt")

from flash import Trainerfrom flash.core.classification import FiftyOneLabelsfrom flash.core.integrations.fiftyone import visualizefrom flash.video import VideoClassificationData, VideoClassifierclassifier = VideoClassifier.load_from_checkpoint(...)# Option 1: Generate predictions using a Trainer and datamoduledatamodule = VideoClassificationData.from_folders(predict_folder="/path/to/folder",...)trainer = Trainer()classifier.serializer = FiftyOneLabels(return_filepath=True)predictions = trainer.predict(classifier, datamodule=datamodule)session = visualize(predictions) # Launch FiftyOne# Option 2: Generate predictions from model using filepathsfilepaths = ["list", "of", "filepaths"]predictions = classifier.predict(filepaths)classifier.serializer = FiftyOneLabels()session = visualize(predictions, filepaths=filepaths) # Launch FiftyOne

pip install fiftyone lightning-flash

pip install umap-learn

from itertools import chainimport fiftyone as foimport fiftyone.zoo as fozfrom flash import Trainerfrom flash.image import ObjectDetectionData, ObjectDetectorfrom flash.image.detection.serialization import FiftyOneDetectionLabels# 1. Load your FiftyOne dataset# Here we use views into one dataset, but you can also create a# different dataset for each splitdataset = foz.load_zoo_dataset("quickstart", max_samples=40)train_dataset = dataset.shuffle(seed=51)[:20]test_dataset = dataset.shuffle(seed=51)[20:25]val_dataset = dataset.shuffle(seed=51)[25:30]predict_dataset = dataset.shuffle(seed=51)[30:40]# 2. Load the Datamoduledatamodule = ObjectDetectionData.from_fiftyone(train_dataset = train_dataset,test_dataset = test_dataset,val_dataset = val_dataset,predict_dataset = predict_dataset,label_field = "ground_truth",batch_size=4,num_workers=4,)# 3. Build the modelmodel = ObjectDetector(model="retinanet",num_classes=datamodule.num_classes,serializer=FiftyOneDetectionLabels(),)# 4. Create the trainertrainer = Trainer(max_epochs=1, limit_train_batches=1, limit_val_batches=1)# 5. Finetune the modeltrainer.finetune(model, datamodule=datamodule)# 6. Save it!trainer.save_checkpoint("object_detection_model.pt")# 7. Generate predictionsmodel = ObjectDetector.load_from_checkpoint("https://flash-weights.s3.amazonaws.com/object_detection_model.pt")model.serializer = FiftyOneDetectionLabels()predictions = trainer.predict(model, datamodule=datamodule)predictions = list(chain.from_iterable(predictions)) # flatten batches# 8. Add predictions to dataset and analyzepredict_dataset.set_values("flash_predictions", predictions)session = fo.launch_app(view=predict_dataset)