Image Classification (ResNet-50)

Training Parameters and Value Ranges

Table 1 describes the names, types, value ranges, default values, and descriptions of the ResNet-50 training parameters.

**Table 1** Training parameters (**resnet50_mindspore**/**model_train.py**)
Parameter	Type	Value Range	Default Value	Description
--train_dataset_path	String	-	None	Path of the training dataset.
--pretrained_ckpt_path	String	-	"pre_trained_ckpt"	Path for storing the pre-trained model.
--train_output_path	String	-	"train_output_path"	Output path of the training result.
--device_num	Integer	Currently, only one device is supported, that is, the value can only be 1.	1	Number of NPUs used for training.
--device_id	Integer	[0,7]	0	ID of the NPU used for training.
--run_eval	Bool	True or False	True	Whether to perform evaluation during training.
--batch_size	Integer	If run_eval is set to False, the value is less than or equal to the number of images in the dataset. If run_eval is set to True, the value is less than or equal to 20% of the number of images in the dataset. [1,8192]	16	Batch size for training.
--epoch_size	Integer	[1,10000]	10	Number of training epochs.
--init_lr	Float	(0,1)	0.0001	Training learning rate.
--enable_modelarts	Bool	True or False	False	Whether to use ModelArts for training.
--data_url	String	-	None	Dataset path for the ModelArts training.
--train_output_url	String	-	None	Output path for the ModelArts training.
--checkpoint_url	String	-	None	Path of the pre-trained model for the ModelArts training.
--net_width	Integer	An integer multiple of 16 within [64,608].	224	Network width.
--net_height	Integer	An integer multiple of 16 within [64,608].	224	Network height.

Training Command Reference for ResNet-50

The command for training the ResNet-50 model is as follows:

python3 model_train.py --train_dataset_path={train_dataset_path} --train_output_path=./output_path --epoch_size=10 --batch_size=16 --init_lr=0.0001 --device_num=1 --device_id=0 --run_eval=True

The model training process is random. Use the evaluated accuracy. Figure 1 shows the training accuracy results.

Figure 1 Training accuracy results of the ResNet-50 model

Figure 2 shows the log information after the model training is complete.

Figure 2 Log information of the ResNet-50 model training

After the training is complete, the .ckpt, .a310.om, .a310p.om, and .air model files are generated in the output directory specified by the --train_output_path parameter.

Evaluation Parameters and Value Ranges

Table 2 describes the names, types, value ranges, default values, and descriptions of the ResNet-50 evaluation parameters.

**Table 2** Evaluation parameters (**resnet50_mindspore**/**model_eval.py**)
Parameter	Type	Value Range	Default Value	Description
--eval_dataset_path	String	-	None	Path of the evaluation dataset.
--eval_ckpt_path	String	-	None	Path for obtaining the evaluation CKPT.
--eval_output_path	String	-	"eval_output_path"	Output path of the evaluation results.
--device_id	Integer	[0,7]	0	ID of the NPU for evaluation.
--enable_modelarts	Bool	True or False	False	Whether to use ModelArts for training.
--data_url	String	-	None	Dataset path for the ModelArts training.
--eval_output_url	String	-	None	Output path for the ModelArts training.
--checkpoint_url	String	-	None	Path of the pre-trained model for the ModelArts training.

Evaluation Command Reference

Run the following command to evaluate the ResNet-50 model:

python3 model_eval.py --eval_dataset_path={eval_dataset_path} --eval_ckpt_path=./output_ path --eval_output_path=./eval_result --device_id=0

Use the checkpoint output during the training to evaluate the model accuracy. Figure 3 shows the results.

Figure 3 Evaluation accuracy results of the ResNet50 model

The folders and file shown in Figure 4 are generated in the evaluation directory. The ng_images folder stores each image that is incorrectly evaluated, the ok_images folder stores each image that is correctly evaluated, and the statistics.csv file stores the evaluation results, including the class name, accuracy rate, recall rate, and accuracy value, as well as the number of tags, detected images and correctly detected images.

Figure 4 Folders and files generated during the ResNet-50 model evaluation

Parent topic: Training Component