metadata

license: apache-2.0
base_model: google/vit-base-patch16-224-in21k
tags:
  - generated_from_trainer
datasets:
  - FastJobs/Visual_Emotional_Analysis
metrics:
  - accuracy
  - precision
  - f1
model-index:
  - name: emo-vit-base-patch16-224-in21k
    results:
      - task:
          name: Image Classification
          type: image-classification
        dataset:
          name: FastJobs/Visual_Emotional_Analysis
          type: FastJobs/Visual_Emotional_Analysis
          config: default
          split: train
          args: default
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.61875
          - name: Precision
            type: precision
            value: 0.6229001976284585
          - name: F1
            type: f1
            value: 0.6163114517061885

emo-vit-base-patch16-224-in21k

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the FastJobs/Visual_Emotional_Analysis dataset. It achieves the following results on the evaluation set:

Loss: 1.2392
Accuracy: 0.6188
Precision: 0.6229
F1: 0.6163

Training and evaluation data

Data Split

Used a 4:1 ratio for training and development sets and a seed of 42.

Pre-processing Augmentation

The main pre-processing phase for both training and evaluation includes:

Resizing to (224, 224, 3)
Normalizing images using a mean and standard deviation of [0.5, 0.5, 0.5]

Other than the aforementioned pre-processing, the training set was augmented using:

Random horizontal & vertical flip
Color jitter
Random resized crop

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 64
eval_batch_size: 64
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine_with_restarts
lr_scheduler_warmup_steps: 10
num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	F1
2.0652	1.0	10	1.9712	0.35	0.3441	0.3294
1.9006	2.0	20	1.6055	0.425	0.3497	0.3578
1.6274	3.0	30	1.4991	0.4875	0.5747	0.4621
1.4742	4.0	40	1.4417	0.4313	0.4744	0.4037
1.3546	5.0	50	1.3699	0.4125	0.3896	0.3387
1.2574	6.0	60	1.2200	0.5125	0.5072	0.4783
1.183	7.0	70	1.1368	0.5375	0.5802	0.5341
1.0869	8.0	80	1.1332	0.5687	0.6024	0.5622
1.002	9.0	90	1.1178	0.55	0.5663	0.5423
0.9453	10.0	100	1.1601	0.5563	0.5994	0.5515
0.9495	11.0	110	1.1202	0.525	0.5695	0.5266
0.7805	12.0	120	1.1620	0.5375	0.5577	0.5323
0.7487	13.0	130	1.2094	0.5687	0.6218	0.5716
0.6805	14.0	140	1.2662	0.5437	0.5875	0.5345
0.6491	15.0	150	1.1673	0.5625	0.5707	0.5511
0.6168	16.0	160	1.2981	0.475	0.5388	0.4846
0.5512	17.0	170	1.2624	0.575	0.6110	0.5726
0.5532	18.0	180	1.2392	0.6188	0.6229	0.6163
0.4931	19.0	190	1.4012	0.5375	0.5542	0.5277
0.4919	20.0	200	1.2323	0.5813	0.5825	0.5758
0.4243	21.0	210	1.3046	0.5875	0.5967	0.5750
0.3971	22.0	220	1.3169	0.5687	0.5812	0.5610
0.3534	23.0	230	1.4052	0.5625	0.6240	0.5527
0.3456	24.0	240	1.3372	0.5875	0.5998	0.5838
0.3381	25.0	250	1.4000	0.55	0.5589	0.5468
0.3786	26.0	260	1.3531	0.5687	0.6269	0.5764
0.3614	27.0	270	1.3696	0.5687	0.6019	0.5704
0.312	28.0	280	1.3523	0.6125	0.6351	0.6148
0.2643	29.0	290	1.4510	0.5813	0.6286	0.5825
0.3553	30.0	300	1.5255	0.6062	0.6560	0.6113
0.2807	31.0	310	1.5901	0.5813	0.5921	0.5655
0.3252	32.0	320	1.5669	0.575	0.5764	0.5639
0.3796	33.0	330	1.6251	0.5375	0.5776	0.5431
0.2635	34.0	340	1.7397	0.4938	0.5513	0.4944
0.2583	35.0	350	1.4806	0.6	0.6566	0.6099
0.3006	36.0	360	1.4808	0.5813	0.6310	0.5863
0.3082	37.0	370	1.7077	0.5188	0.5680	0.5156
0.3346	38.0	380	1.6861	0.575	0.6725	0.5638
0.291	39.0	390	1.5484	0.5625	0.5631	0.5535
0.2313	40.0	400	1.4933	0.5563	0.5564	0.5526
0.2163	41.0	410	1.5836	0.5938	0.6046	0.5929
0.2201	42.0	420	1.6363	0.5687	0.5954	0.5672
0.2077	43.0	430	1.6746	0.5687	0.5623	0.5622

Framework versions

Transformers 4.33.1
Pytorch 2.0.1+cu118
Datasets 2.14.5
Tokenizers 0.13.3