B. Catboost Regressor Model Building
Step 14: Train-Test-Valid Split for Regression:
In the below code we do a train-test-valid split. We have 70,168 records for training, 30,072 records for validation and 33,414 records for testing. Set a random state to ensure model output repeatability.
Python3
x = df[ 'image' ] y = df[ 'cloudcover' ] x_train, x_test, y_train, y_test = train_test_split(x, y, test_size = 0.25 , random_state = 48 ) x_train, x_val, y_train, y_val = train_test_split(x_train, y_train, test_size = 0.30 , random_state = 48 ) print ((x_train.shape, x_val.shape, x_test.shape)) |
Output:
((70168,), (30072,), (33414,))
Step 15: Loading the Finetuned Clip Model:
In the below code we load the finetuned .pt CLIP model. Now since we have our finetuned CLIP model which is ready for image feature extraction, we first of all extract the feature embeddings for all the sky cam images from CLIP model and those embeddings would work as input features for our catboost model.
Python3
model = CLIPModel().to(CFG.device) model.load_state_dict(torch.load( "clip_model.pt" , map_location = CFG.device)) model. eval () |
Output:
CLIPModel(
(image_encoder): ImageEncoder(
(model): ResNet(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(layer1): Sequential(
(0): Bottleneck(
(conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
)
(layer2): Sequential(
(0): Bottleneck(
(conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
)
(layer3): Sequential(
(0): Bottleneck(
(conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(3): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(4): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(5): Bottleneck(
(conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
)
(layer4): Sequential(
(0): Bottleneck(
(conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
(downsample): Sequential(
(0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Bottleneck(
(conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
(2): Bottleneck(
(conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act1): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(drop_block): Identity()
(act2): ReLU(inplace=True)
(aa): Identity()
(conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
(bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(act3): ReLU(inplace=True)
)
)
(global_pool): SelectAdaptivePool2d (pool_type=avg, flatten=Flatten(start_dim=1, end_dim=-1))
(fc): Identity()
)
)
(text_encoder): TextEncoder(
(model): DistilBertModel(
(embeddings): Embeddings(
(word_embeddings): Embedding(30522, 768, padding_idx=0)
(position_embeddings): Embedding(512, 768)
(LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
(dropout): Dropout(p=0.1, inplace=False)
)
(transformer): Transformer(
(layer): ModuleList(
(0-5): 6 x TransformerBlock(
(attention): MultiHeadSelfAttention(
(dropout): Dropout(p=0.1, inplace=False)
(q_lin): Linear(in_features=768, out_features=768, bias=True)
(k_lin): Linear(in_features=768, out_features=768, bias=True)
(v_lin): Linear(in_features=768, out_features=768, bias=True)
(out_lin): Linear(in_features=768, out_features=768, bias=True)
)
(sa_layer_norm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
(ffn): FFN(
(dropout): Dropout(p=0.1, inplace=False)
(lin1): Linear(in_features=768, out_features=3072, bias=True)
(lin2): Linear(in_features=3072, out_features=768, bias=True)
(activation): GELUActivation()
)
(output_layer_norm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
)
)
)
)
)
(image_projection): ProjectionHead(
(projection): Linear(in_features=2048, out_features=256, bias=True)
(gelu): GELU(approximate='none')
(fc): Linear(in_features=256, out_features=256, bias=True)
(dropout): Dropout(p=0.1, inplace=False)
(layer_norm): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
)
(text_projection): ProjectionHead(
(projection): Linear(in_features=768, out_features=256, bias=True)
(gelu): GELU(approximate='none')
(fc): Linear(in_features=256, out_features=256, bias=True)
(dropout): Dropout(p=0.1, inplace=False)
(layer_norm): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
)
)
Step 16: Dataset Preparation for Regression Model:
In the below code, we are creating a custom Torch dataset for our images and label. We load all the images and labels to this dataset which will be passed to the finetuned model for predictions.
Python3
# ----- Custom Dataset Loader ----- # class SkyImage(Dataset): def __init__( self , img_dir, labels): self .img_dir = img_dir self .img_labels = labels def __len__( self ): return len ( self .img_dir) def __getitem__( self , idx): img_path = os.path.join(img_folder, self .img_dir[idx]) #os.path.join("Extracted Images/", self.img_dir[idx]) image = cv2.imread(img_path) image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) image = cv2.resize(image, ( 244 , 244 )) image = np.moveaxis(image, - 1 , 0 ) label = self .img_labels[idx] return image, label # ----- Dataset ----- # train_images = SkyImage(x_train.to_list(), y_train.to_list()) valid_images = SkyImage(x_val.to_list(), y_val.to_list()) test_images = SkyImage(x_test.to_list(), y_test.to_list()) |
Step 17: Extracting image features from CLIP Model:
Check if CUDA (GPU) is available and set the device accordingly
Python3
# Check if CUDA (GPU) is available and set the device accordingly device = torch.device( "cuda" if torch.cuda.is_available() else "cpu" ) device |
Output:
device(type='cpu')
Building Features
The below get features method intakes our torch dataset for training, testing & validation set and extract image feature vectors from skycam images and returns the extracted feature vectors.
Python3
# ----- Building Features ----- # def get_features(dataset): all_features, all_labels, all_embeddings = [], [], [] with torch.no_grad(): for images, labels in tqdm(DataLoader(dataset, batch_size = 64 )): image_input = torch.tensor(np.stack(images)).to(device). float () image_features = model.image_encoder(image_input) image_embeddings = model.image_projection(image_features) all_features.append(image_features) all_labels.append(labels) all_embeddings.append(image_embeddings) return torch.cat(all_features), torch.cat(all_labels).to(device), torch.cat(all_embeddings).to(device) # ----- Get Features ----- # train_features, train_labels, train_embeddings = get_features(train_images) valid_features, valid_labels, valid_embeddings = get_features(valid_images) test_features, test_labels, test_embeddings = get_features(test_images) |
Step 18: Evaluation Metrics Method:
We build our custom metrics method to evaluate our catboost model. We will be using Mean Absolute Error (MAE), Root Mean Square Error (RMSE) & R-Squared (R2) as our model evaluators.
Python3
def evaluate(name, x, y, n, p): # p: features, #n: no of observations print ( "---------------------------------------------------" ) print ( "{} MAE: {}" . format (name, mean_absolute_error(x, y))) print ( "{} RMSE: {}" . format (name, mean_squared_error(x, y, squared = False ))) print ( "{} MSE: {}" . format (name, mean_squared_error(x, y))) r2 = r2_score(x, y) print ( "{} R2: {}" . format (name, r2)) print ( "---------------------------------------------------" ) |
Step 19: Catboost model Training:
We train the catboost model with multiple hyperparameters. Lets use RMSE as evaluation metrics to evaluate model on validation data. We pass both training & validation data to fit method. By this way we ensure that there is no model overfitting because model is validation on validation data which is not present in training data. I have found out the best hyperparameters for this scenario and have used those in below code. For finding out best hyperparameters you can do hit and trial or apply cross validation.
Python3
# ----- Model Training ----- # CB_model = CatBoostRegressor(iterations = 700 , learning_rate = 0.1 , max_depth = 8 , eval_metric = 'RMSE' , random_seed = 48 ) CB_model.fit(train_features.cpu().numpy(), train_labels.cpu().numpy(), eval_set = (valid_features.cpu().numpy(), valid_labels.cpu().numpy()), use_best_model = True , plot = True , verbose = 50 ) |
Output:
0: learn: 28.1361841 test: 28.2423136 best: 28.2423136 (0) total: 2.13s remaining: 24m 49s
50: learn: 11.5614561 test: 11.9335237 best: 11.9335237 (50) total: 1m 3s remaining: 13m 21s
100: learn: 10.7263689 test: 11.4059249 best: 11.4059249 (100) total: 2m 1s remaining: 12m 1s
150: learn: 10.0566562 test: 11.0617557 best: 11.0617557 (150) total: 3m remaining: 10m 55s
200: learn: 9.5172739 test: 10.8473396 best: 10.8473396 (200) total: 3m 58s remaining: 9m 51s
250: learn: 9.0923719 test: 10.6886373 best: 10.6886373 (250) total: 4m 55s remaining: 8m 47s
300: learn: 8.7042622 test: 10.5734544 best: 10.5734544 (300) total: 5m 51s remaining: 7m 45s
350: learn: 8.3755575 test: 10.4773273 best: 10.4773273 (350) total: 6m 47s remaining: 6m 45s
400: learn: 8.0759744 test: 10.3938604 best: 10.3938604 (400) total: 7m 44s remaining: 5m 46s
450: learn: 7.7814581 test: 10.3233375 best: 10.3233375 (450) total: 8m 42s remaining: 4m 48s
500: learn: 7.5160766 test: 10.2628795 best: 10.2628795 (500) total: 9m 39s remaining: 3m 50s
550: learn: 7.2897423 test: 10.2027638 best: 10.2027638 (550) total: 10m 35s remaining: 2m 51s
600: learn: 7.0611325 test: 10.1574324 best: 10.1574324 (600) total: 11m 33s remaining: 1m 54s
650: learn: 6.8320990 test: 10.1136860 best: 10.1136860 (650) total: 12m 30s remaining: 56.5s
699: learn: 6.6529638 test: 10.0780409 best: 10.0780409 (699) total: 13m 25s remaining: 0us
bestTest = 10.07804086
bestIteration = 699
Step 20: Model Prediction
Lets do the prediction on Train, Test & Validation sets.
Python3
# ----- Model Prediction ----- # cbt_train_pred = CB_model.predict(train_features.cpu().numpy()) cbt_valid_pred = CB_model.predict(valid_features.cpu().numpy()) cbt_test_pred = CB_model.predict(test_features.cpu().numpy()) |
Step 21: Model Evaluation
Now, we have the actual values of cloud cover and predicted values of cloud cover we can evaluate our Catboost model.
Python3
# ----- Model Evaluation ----- # evaluate( "Train" , train_labels.cpu(), cbt_train_pred, len (cbt_train_pred), 1 ) evaluate( "Valid" , valid_labels.cpu(), cbt_valid_pred, len (cbt_valid_pred), 1 ) evaluate( "Test" , test_labels.cpu(), cbt_test_pred, len (cbt_test_pred), 1 ) |
Output:
---------------------------------------------------
Train MAE: 4.437975369402876
Train RMSE: 6.652963762088708
Train MSE: 44.26192681966554
Train R2: 0.9523583786704957
---------------------------------------------------
---------------------------------------------------
Valid MAE: 6.304070193782646
Valid RMSE: 10.078040861839906
Valid MSE: 101.56690761291485
Valid R2: 0.8914442298156392
---------------------------------------------------
---------------------------------------------------
Test MAE: 6.364711156454016
Test RMSE: 10.198410458657648
Test MSE: 104.0075758832577
Test R2: 0.889060898998321
---------------------------------------------------
Insights:
- Out model is perfectly fitted on the data, and can be deployed on UI.
- Test & Validation Metrics are very good.
- There is no overfitting as the Train Metrics are closer to Validation & Test Metrics.
Step 22: Save the Catboost Model
We save the Catboost Model for deploying purpose.
Python3
pickle.dump(CB_model, open ( 'catboost_model.sav' , 'wb' )) |
Cloud Coverage Prediction using Skycam Images
Cloud coverage prediction is critical in weather forecasting and a variety of applications such as solar energy generation, aviation, and climate monitoring. Accurate forecasts help decision-makers and sectors plan for and adapt to changing weather conditions. The advancement of artificial intelligence and computer vision techniques in recent years has created new opportunities for enhancing cloud coverage forecasts.
One promising approach is the use of SkyCam images.
- In the face of rapidly changing global climate patterns, there is an urgent need for innovative tools and technologies to better understand and predict weather-related phenomena.
- One crucial aspect of climate analysis is the assessment of cloud coverage, which plays a pivotal role in influencing weather conditions and climate trends.
- Experts may not always be available to monitor climatic shifts. Therefore, developing an automated weather monitoring system is crucial for various applications, including agriculture and disaster management.
The purpose of this research is to estimate the opaque Cloud Coverage from a Skycam Image using AI/ML methodologies.
Table of Content
- Cloud Coverage Prediction using SkyCam Images
- Implementations Cloud Coverage Prediction using SkyCam Images
- Cloud Coverage Prediction Models:
- Part I. Model Building & Traning Pipeline
- A. Clip Model Finetuning
- B. Catboost Regressor Model Building
- Part II. UI Inference Codes for Deployed Model
- Results:
Contact Us