FIGURE 4 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
To solve the problem of the low efficiency of traditional lettuce freshness classification methods and sample damage, we proposed an automatic lettuce freshness classification method based on improved deep residuals convolutional neural network (Im-ResNet). We built an image acquisition system to obtain the freshness classification dataset of lettu...
Contexts in source publication
Context 1
... on this, we constructed a new network model and named it Im-ResNet. The detailed description of Im-ResNet architecture is shown in Figure 4. ...
Context 2
... shown in Figure 4, Im-ResNet model is divided six stages. The network performs a 3 × 3 zero-padded operation to maintain the boundary information of an image with an input size of 224 × 224 × 3. ...
Context 3
... 6: The newly constructed convolutional layer C1, pooling layer P1 and two fully connected layers F1, F2 are shown in the orange dotted box in Figure 4. Here, the filter size of convolutional layer C1 is 3 × 3 pixels and S is set to 3. The total number of filters is 2048. ...
Context 4
... order to improve the classification accuracy, Im-ResNet was trained six times on the training set. First, all stages of the model were frozen, and only C1, P1 and F1 and F2 in VOLUME 10, 2022 the orange dashed boxes were trained, as shown in Figure 4. Then the last convolutional block was unfrozen to retrain the network, and only stage 5 was unfrozen. ...
Similar publications
Hammer testing, a nondestructive testing method, has been demonstrated to provide information on structural damage. One of the biggest challenges with this testing method is the simultaneous identification of surface, internal, and composite damage (consisting of both surface and internal damage) in a complex environment, such as post-disaster. A m...
Citations
... In the same way, Wang used the Grad-CAM method to visualize the results of freshness level classification, and the results indicated that the abdominal features of the shrimp were the most effective information for freshness grading [66]. Besides, Xu also used Grad-CAM to accurately extract classification features and accurately identify lettuce leaves of four freshness levels, including those that withered and became smaller [68]. These studies were sufficient to prove that in the field of image classification, the Grad-CAM method could accurately find the classification features and achieve accurate classification. ...
Conventional evaluation of fish freshness based on physiological and biochemical methods was destructive, complicated and costly. In this study, the new model was trained on the eye images of 100 large yellow croakers along with their total volatile basic nitrogen (TVB-N) value as freshness indicators in the storage of nine consecutive days at 4 °C. The experiment was divided into three stages (0–2 days, 3–6 days, and 7–8 days) based on TVB-N value, about 1000 images in each stage were used for freshness classification. A non-destructive and rapid fish freshness detection method based on the eye region images of large yellow croaker was proposed by mathematical modeling. The features of large yellow croaker images were extracted automatically by ResNet-34 structure, and then the key extracted feature was focused on the pupil of the fish eye by mixed attention mechanism. Finally, the features of pupil were used to classify the freshness of large yellow croaker. The results showed the accuracy of the model to classify the fish freshness was reached to 99.4%. The model constructed based on the eye images was non-destructive, and could well monitor and distinguish the freshness of large yellow croakers at different storage stages.
... Here, the accuracy is 97.50%. [10] suggested an automatic lettuce freshness classification approach based on enhanced deep residuals CNN to address the issue of sample damage and the ineffectiveness of conventional lettuce freshness classification methods (Im-ResNet). Utilizing the network in comparison to four other network topologies, XU et al. carried out the associated tests (AlexNet, GoogleNet, VGG16 and ResNet50) that demonstrated that the suggested network had more notable benefits in recognition accuracy (95.60%) and least loss value of lettuce freshness. ...
A vegetable’s quality performs a significant role
in customer consumption. At the same time, the categorization
of vegetable freshness is crucial for the food industry. Freshness
is a key indicator of vegetable quality that has a direct impact
on human physical well-being and desire to make purchases.
Initially, identifying the freshness condition of vegetables and
distinguishing among fresh, aged, and rotten vegetables by the
observation of vegetable’s outer shell manually is very difficult
for humans. That can be eradicated by replacing the monitoring
system with an automated computer program. An automatic
fresh vegetable detection system is proposed using the
Densenet201 Transfer Learning model in this study. The
primary goal of this research is to identify a vegetable’s
freshness condition by observing the outer shell and
differentiate a fresh vegetable from a rotten one. Five types of
vegetables are divided into three classes using custom datasets
for vegetable freshness classification using different transfer
learning models. However, DenseNet201 performed enormously
enough on the vegetable dataset which achieved a test accuracy
of 98.56%. Thus, this study will attempt to assist in reducing the
reliance on the human eye and accurately identify the freshness
conditions of a vegetable.
... RReLU is a variant of Leaky ReLU [42]. In RReLU, the slope of the negative values is random in training and becomes fixed in testing. ...
Nowadays, tourists increasingly prefer to check the reviews of attractions before traveling to decide whether to visit them or not. To respond to the change in the way tourists choose attractions, it is important to classify the reviews of attractions with high precision. In addition, more and more tourists like to use emojis to express their satisfaction or dissatisfaction with the attractions. In this paper, we built a dataset for Chinese attraction evaluation incorporating emojis (CAEIE) and proposed an explicitly n-gram masking method to enhance the integration of coarse-grained information into a pre-training (ERNIE-Gram) and Text Graph Convolutional Network (textGCN) (E2G) model to classify the dataset with a high accuracy. The E2G preprocesses the text and feeds it to ERNIE-Gram and TextGCN. ERNIE-Gram was trained using its unique mask mechanism to obtain the final probabilities. TextGCN used the dataset to construct heterogeneous graphs with comment text and words, which were trained to obtain a representation of the document output category probabilities. The two probabilities were calculated to obtain the final results. To demonstrate the validity of the E2G model, this paper was compared with advanced models. After experiments, it was shown that E2G had a good classification effect on the CAEIE dataset, and the accuracy of classification was up to 97.37%. Furthermore, the accuracy of E2G was 1.37% and 1.35% ahead of ERNIE-Gram and TextGCN, respectively. In addition, two sets of comparison experiments were conducted to verify the performance of TextGCN and TextGAT on the CAEIE dataset. The final results showed that ERNIE and ERNIE-Gram combined TextGCN and TextGAT, respectively, and TextGCN performed 1.6% and 2.15% ahead. This paper compared the effects of eight activation functions on the second layer of the TextGCN and the activation-function-rectified linear unit 6 (RELU6) with the best results based on experiments.
Freshness is a prime factor of consideration when purchasing consumables like fruits and vegetables. Studies have proven that Computer Vision can be successfully involved in classifying fresh and stale fruits and vegetables and measuring their freshness to some extent. This work attempts to determine and analyze the freshness of fruits and vegetables from their images by proposing a Machine Learning methodology. The entire study was divided into two steps. The first step focused on obtaining classification between images of fresh and stale fruits and vegetables. For this, we trained the ConvNeXt model on an open-source imagery dataset consisting of 12 classes, and it proved efficient by achieving an accuracy of 99.77%. The second step focused on analyzing how fresh a particular fruit or vegetable is from its image. We achieved this by using an open-source dataset of tomato images and extracting features specific to the texture, shape, and colour from these images. Further, we trained classification models on these extracted features and presented the results as quantitative measures with scores of 10for each of these three factors. Thus, we attempted to achieve an in-depth freshness analysis by grading the images based on these three critical factors while defining freshness.
