Strawberries, known for their economic significance and rich nutritional value, are cultivated extensively worldwide. However, a host of workers need to be employed every year to identify and categorize the developmental stages of the strawberries in the greenhouses, which is not only time-consuming, inefficient, increasing the cultivation cost, but also difficult to guarantee the classification accuracy. Meanwhile, affected by the complicated background, occlusions, and color interference, the features of strawberries are proven challenging to be extracted via the traditional neural networks due to serious gradient disappearance. Therefore, an improved CBAM-ResNet34-based classification evaluation method for developmental processes of greenhouse strawberries is investigated. The procedure of this method is as follows: firstly, the developmental stages of greenhouse strawberries are classified by experts into four stages: Stage I (initial stage), Stage II (green and white fruit stage), Stage III (early ripening stage), and Stage IV (fully ripe stage). The 627, 640, 604, and 340 strawberry images for these four stages are captured. Subsequently, the images are divided into training, validation, as well as testing sets and then undergo image pre-processing, expansion, and augmentation. Whereafter, the 7x7 convolution kernel in the first layer of the network is replaced by three consecutive 3x3 convolution cores to eliminate the redundant weights and unnecessary model parameters, and the BasicBlocks configuration is adjusted. Finally, the CBAM attention mechanism is added to each BasicBlock so as to pinpoint the spatial position of the strawberries and extract their major features such as shape, size, and color. Comparison experiments with the conventional deep neural networks LeNet5, AlexNet, VGG16, ResNet18, ResNet34, and every improved part of CBAM-ResNet34 demonstrated that when the learning rate is 0.001, the Dropout rate is 0.3, and the Adam's weight decay parameter is 0.001, the accuracies for validation and testing sets can reach to 92.36% and 87.56% with F1 scores of 0.92, 0.87, 0.85 and 0.88.