How do I reduce 4096-dimensional feature vector to 1024-dimensional vector in CNN Caffemodel?

Question

I used 16-layers VGGnet to extract features from an image. It outputs a 4096-dimensional feature vector. However, I need a 1024-dimensional vector. How do I further reduce this 4096-vector into 1024-vector? Do I need to add a new layer on top of fc7?

If you simply add another layer that layer's weights wont be trained. So your feature vector might just be random. — DollarAkshay, Jul 02 '18 at 08:10

score 3 · Answer 1 · answered Mar 02 '16 at 11:16

Yes, you need to add another layer on top of fc7. This is how your last few layers should be like

layers {
  bottom: "fc7"
  top: "fc7"
  name: "relu7"
  type: RELU
}
layers {
  bottom: "fc7"
  top: "fc7"
  name: "drop7"
  type: DROPOUT
  dropout_param {
    dropout_ratio: 0.5
  }
}
layers {
  name: "fc8"
  bottom: "fc7"
  top: "fc8"
  type: INNER_PRODUCT
  inner_product_param {
    num_output: 1024
  }
  blobs_lr: 0
  blobs_lr: 0
}
layers {
  name: "loss"
  type: SOFTMAX_LOSS
  bottom: "fc8"
  bottom: "label"
  top: "loss/loss"
}
layers {
  name: "accuracy/top1"
  type: ACCURACY
  bottom: "fc8"
  bottom: "label"
  top: "accuracy@1"
  include: { phase: TEST }
  accuracy_param {
    top_k: 1
  }
}

score 2 · Answer 2 · answered Jan 06 '16 at 06:28

2

Yes.‏‏‏ ‏‏ ‏‏‏‏‏‏‏‏‏‏‏‏‏‏‏‏‏‏‏

answered Jan 06 '16 at 06:28

Shai

111,146
38
238
371

@Shai will you please elaborate ?? – Omkar Acharya Jan 06 '16 at 07:47
1

@foo well you asked if you need to add another layer on top of `fc7` to reduce output dim from 4096 to 1024, and the answer is YES that's the elegant way of reducing dimension trained for the specific task you are trying to achieve. – Shai Jan 06 '16 at 07:49
2

This does not provide an answer to the question. To critique or request clarification from an author, leave a comment below their post. - [From Review](/review/low-quality-posts/10790862) – Andrew Stephens Jan 06 '16 at 09:13
4

@AndrewStephens if you read the question and my comment above you. you'll see that it does answer – Shai Jan 06 '16 at 10:52

How do I reduce 4096-dimensional feature vector to 1024-dimensional vector in CNN Caffemodel?

2 Answers2

Linked