/
4_Tomographic_Volumetric_Additive_Manufacturing_with_Refraction.jl
358 lines (282 loc) · 12.6 KB
/
4_Tomographic_Volumetric_Additive_Manufacturing_with_Refraction.jl
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
### A Pluto.jl notebook ###
# v0.19.38
using Markdown
using InteractiveUtils
# This Pluto notebook uses @bind for interactivity. When running this notebook outside of Pluto, the following 'mock version' of @bind gives bound variables a default value (instead of an error).
macro bind(def, element)
quote
local iv = try Base.loaded_modules[Base.PkgId(Base.UUID("6e696c72-6542-2067-7265-42206c756150"), "AbstractPlutoDingetjes")].Bonds.initial_value catch; b -> missing; end
local el = $(esc(element))
global $(esc(def)) = Core.applicable(Base.get, el) ? Base.get(el) : iv(el)
el
end
end
# ╔═╡ 179d107e-b3be-11ee-0a6c-49f1bf2a10fd
begin
using Pkg
Pkg.activate(".")
Pkg.instantiate()
using Revise
end
# ╔═╡ e3a3dc83-6536-445f-b56b-d8ad62cd9a85
using TestImages, RadonKA, ImageShow, ImageIO, Noise, PlutoUI, BenchmarkTools, CUDA, Zygote, IndexFunArrays, FileIO, NDTools, Plots, ProgressLogging, Optim
# ╔═╡ eb4b7177-4f93-4728-8c91-4ff6a86f40cf
md"# Load packages and check CUDA"
# ╔═╡ f3277f00-4162-43b2-98e0-53bcb508df1e
TableOfContents()
# ╔═╡ 42f31ceb-0f5d-46cc-b868-0314b41a5407
begin
use_CUDA = Ref(CUDA.functional())
var"@mytime" = use_CUDA[] ? CUDA.var"@time" : var"@time"
togoc(x) = use_CUDA[] ? CuArray(x) : x
togoc(x::AbstractArray{Float64}) = use_CUDA[] ? CuArray(Float32.(x)) : x
end
# ╔═╡ 78746412-00f2-40ad-9d24-e884d8dd7e77
md"""# CUDA is functional: $(CUDA.functional() ? "yes" : "no")"""
# ╔═╡ 980e9656-202d-4a47-837f-b2197fe5d3b4
md"# Load JOSS logo as target"
# ╔═╡ f0d86933-5eab-4d86-8909-c51ab9ca99d9
background = select_region(ImageShow.Gray.(load(download("https://upload.wikimedia.org/wikipedia/commons/thumb/9/94/JOSS.png/459px-JOSS.png"))) .< 0.001, new_size=(660,660));
# ╔═╡ 36eab63f-5643-4b43-8e12-0f7d971c1040
img = select_region(.-ImageShow.Gray.(load(download("https://upload.wikimedia.org/wikipedia/commons/thumb/9/94/JOSS.png/459px-JOSS.png"))), new_size=(660,660));
# ╔═╡ fc049fbc-4ae7-4256-b39f-27fdff98d45a
target = togoc(Float32.((img .- background) .> 0.1));
# ╔═╡ a6bdc90f-3c10-4524-9892-6e976d556a95
simshow(Array(target))
# ╔═╡ 066664f7-9ae5-4a4d-b541-663067cea361
md"# Specify Angles and ray endpoints
We consider that we have a vial with outer radius 7mm and a inner radius of 6.5mm. The vial has a refractive index of 1.5 whereas the resin has one of 1.48.
The full derivation of `distort_rays_vial` is linked in this [PDF](https://github.com/roflmaostc/RadonKA.jl/blob/main/docs/src/angle_derivation/refraction_inside_vial_resin.tex).
"
# ╔═╡ 5ef75bb5-5ab0-4e98-9026-a8e2c5b7833d
angles = togoc(range(0, 2f0 * π, 400));
# ╔═╡ 5aa03590-c552-410e-93aa-1e47dabd2f9e
N_s = size(target,1) - 1
# ╔═╡ 9d582f0d-7da4-41db-9369-fef7a6dbe590
range(-N_s/2 - 1, N_s / 2 - 1, N_s)
# ╔═╡ 6b91d2bd-1170-4ce5-bb20-10d69f8cd6d7
md"
The `ray_startpoints` and `ray_endpoints` indicate each the beginning and end position of the ray.
"
# ╔═╡ 8f212882-5a4f-4915-a01a-9c7f342ea1ef
Rvial = 7f-3
# ╔═╡ 07ff2197-836d-4e9e-b716-4e0eacd89e6e
Rresin = 6.5f-3
# ╔═╡ 6deeaf05-1008-4f1e-bb33-2b74196c9d80
nvial = 1.5f0
# ╔═╡ b94f82e8-1c24-46a0-b3e9-f98a0e3e1c12
nresin = 1.48f0
# ╔═╡ 66a09fd3-545b-4456-be74-6593d2d15a28
radius_pixel = N_s/2f0
# ╔═╡ af83cb81-49b8-41f4-a24e-c1e33e76d925
function distort_rays_vial(y::T, Rₒ::T, Rᵢ::T, nvial::T, nresin::T) where T
y, Rₒ, Rᵢ, nvial, nresin = Float64.((y, Rₒ, Rᵢ, nvial, nresin))
if iszero(y)
return zero(T), zero(T)
end
α = asin(y / Rₒ)
β = asin(sin(α) / nvial)
x = Rₒ * cos(β) - sqrt(Rₒ^2*(cos(β)^2 - 1) + Rᵢ^2)
ϵ = acos((-Rₒ^2 + x^2 + Rᵢ^2) / 2 / Rᵢ / x) - Float64(π) / 2
β_ = sign(y) * (Float64(π) / 2 - ϵ)
γ = asin(nvial * sin(β_) / nresin)
δ₁ = α - β
δ₂ = β_ - γ
δ_ges = δ₁ + δ₂
y_ = abs(Rᵢ * sin(γ))
p = sqrt(Rₒ^2 - y_^2)
η = - (asin(p / Rₒ) - sign(y) * (Float64(π)/2 - δ_ges))
yf = Rₒ * sin(η)
yi = 2 * p * sin(δ_ges) + yf
return T(yi), T(yf)
end
# ╔═╡ 6721b65f-48ed-4dc7-99dd-c460bb124c1e
ray_points = distort_rays_vial.(range(-N_s/2f0 + 1f0, N_s / 2f0 - 1f0, N_s),
radius_pixel,
radius_pixel / Rvial * Rresin,
nvial,
nresin)
# ╔═╡ 76b4c0de-af7a-456e-8d49-04f10a0a17fb
ray_startpoints, ray_endpoints = map(first, ray_points), map(x->x[2], ray_points)
# ╔═╡ 4cdc5592-3b92-4e98-8718-d77dd939ebf1
md"## Ray distribution inside vial"
# ╔═╡ 494c5b20-0af2-4903-959c-812459877c31
kk = 1
# ╔═╡ dc22dead-1b83-4787-a147-9811136af085
begin
sinogram = zeros((length(ray_startpoints), kk))
sinogram[1:1:end,:] .= 1
end;
# ╔═╡ e10549a7-0f31-48fa-8d38-820907d8554e
Revise.errors()
# ╔═╡ e83be634-dde4-4ed4-9226-2ba895a84d15
geometry = RadonKA.RadonFlexibleCircle(size(target, 1), ray_startpoints, ray_endpoints)
# ╔═╡ 85b6e54c-3256-4ba8-9889-56be82de0203
simshow(Array(backproject(togoc(sinogram), togoc(range(0, 2π, kk + 1)[1:end-1]); geometry, μ=0.01)), γ=0.6)
# ╔═╡ 5d55464f-c7d9-4e45-a1e7-907c2137be95
md"# Define Optimization algorithm
See this paper:
[*Charles M. Rackson, Kyle M. Champley, Joseph T. Toombs, Erika J. Fong, Vishal Bansal, Hayden K. Taylor, Maxim Shusteff, Robert R. McLeod, Object-space optimization of tomographic reconstructions for additive manufacturing*](https://www.sciencedirect.com/science/article/abs/pii/S2214860421005212)
"
# ╔═╡ 834cd0f6-9be0-4589-82c0-ca0021911dd5
function make_fg!(fwd_operator, target, threshold)
isobject = target .== 1
notobject = target .== 0
f = let fwd_operator=fwd_operator
# apply sqrt for anscombe
f(x) = begin
fwd = fwd_operator(x)
return (sum(abs2, max.(0, fwd[notobject] .- threshold[1])) +
sum(abs2, max.(0, threshold[2] .- fwd[isobject])) +
sum(abs2, max.(0, fwd[isobject] .- 1)))
end
end
g! = let f=f
g!(G, x) = begin
if !isnothing(G)
G .= Zygote.gradient(f, x)[1]
end
end
end
return f, g!
end
# ╔═╡ 697d4c7d-697b-452c-8a6a-25f2a1089529
function iter!(buffer, img, θs, μ; clip_sinogram=true, geometry)
sinogram = radon(img, θs; μ, geometry)
if clip_sinogram
sinogram .= max.(sinogram, 0)
end
img_recon = backproject(sinogram, θs; μ, geometry)
img_recon ./= maximum(img_recon)
buffer .= max.(img_recon, 0)
return buffer, sinogram
end
# ╔═╡ 624858de-e7da-4e3c-91b5-6ab7b5e645d5
# see https://www.sciencedirect.com/science/article/abs/pii/S2214860421005212
function OSMO(img::AbstractArray{T}, θs, μ=nothing;
thresholds=(0.65, 0.75), N_iter = 2,
geometry) where T
N = size(img, 1)
guess = copy(img)
notobject = togoc(iszero.(img))
isobject = togoc(isone.(img))
losses = T[]
buffer = copy(img)
tmp, s = iter!(buffer, guess, θs, μ; clip_sinogram=true, geometry)
@progress for i in 1:N_iter
guess[notobject] .-= max.(0, tmp[notobject] .- thresholds[1])
tmp, s = iter!(buffer, guess, θs, μ; clip_sinogram=true, geometry)
guess[isobject] .+= max.(0, thresholds[2] .- tmp[isobject])
end
printed = backproject(s, θs; μ, geometry)
printed ./= maximum(printed)
return guess, s, printed
end
# ╔═╡ 12e7c6b7-fcf3-44e4-9d80-785704d93cb5
fwd = x -> backproject(max.(0,x), angles; geometry)
# ╔═╡ 942e1cf7-a47b-43af-a460-1ea896a2c9b8
f, g! = make_fg!(fwd, target, (0.65, 0.75))
# ╔═╡ d80f6c47-1c55-4a7e-9374-104806873a9a
md"# Run optimization
With a decent GPU (such as RTX 3060) it should take around ~20-30s.
With a 12 core multithreaded CPU around ~15min.
"
# ╔═╡ be45bfe1-e91e-45eb-8902-62e8df0d8ac7
begin
init0 = radon(target, angles);
init0 .= 0
end;
# ╔═╡ e14a65c3-ece0-4d1e-96a7-9a9ab00001ad
@time res = Optim.optimize(f, g!, init0, LBFGS(),
Optim.Options(iterations = 15,
store_trace=true))
# ╔═╡ 2ca8698e-3ff8-445c-a5e3-3a20d2e2afc3
@mytime a_object, patterns, printed = OSMO(target, angles, 1/350f0; N_iter=200,
geometry)
# ╔═╡ ec528b63-95eb-430b-b683-04e207dd61f0
simshow((Array(printed) .> 0.7) .- 0 .* Array(target), cmap=:turbo)
# ╔═╡ 43e4b941-b838-483d-9a8f-8897720f8b90
simshow(Array(printed), cmap=:turbo)
# ╔═╡ 45884ebb-a02d-446a-a7ca-05b8b4abc244
simshow(Array(fwd(res.minimizer)), cmap=:turbo)
# ╔═╡ 7c998816-b4de-4dbf-977b-bdab9355be79
simshow(Array(backproject(patterns[:, 96:96], togoc(angles[96:96]); μ=1/350f0, geometry)), cmap=:turbo)
# ╔═╡ 28d8c378-b818-4d61-b9d4-a06ad7cc21e3
simshow(Array(backproject(patterns[:, 306:306], togoc(angles[306:306]); μ=1/350f0, geometry)), cmap=:turbo)
# ╔═╡ beb61bb1-3f7c-4301-92ca-5418bc0445cf
simshow(Array(backproject(patterns[:, 1:1], togoc(angles[1:1]); μ=1/350f0, geometry)), cmap=:turbo)
# ╔═╡ 88895d8d-432e-4337-a51c-12ac4e4a4325
@bind i Slider(1:size(angles,1), show_value=true)
# ╔═╡ 61048e10-8e23-41a6-8c49-0b1a8c0adc24
simshow(Array(backproject(patterns[:, i:i], togoc(angles[i:i]); μ=1/350f0, geometry)), cmap=:turbo)
# ╔═╡ 21c0b369-a4b2-43ea-87b8-702474d88584
simshow(Array(patterns))
# ╔═╡ b53c8016-177f-4b1b-b07d-48cbd2b2fc65
function plot_histogram(target, object_printed, thresholds; yscale=:log10)
# :stephist vs :barhist
plot_font = "Computer Modern"
default(fontfamily=plot_font,
linewidth=2, framestyle=:box, label=nothing, grid=false)
plot(object_printed[target .== 0], seriestype=:stephist, bins=(0.0:0.01:1), xlim=(0.0, 1.0), label="dose distribution void", ylabel="voxel count", xlabel="normalized intensity", ylim=(10, 500000), linewidth=1, legend=:topleft, yscale=yscale, size=(500, 300))
plot!(object_printed[target .== 1], seriestype=:stephist, bins=(0.0:0.01:1), xlim=(0.0, 1.0), label="dose distribution object", ylabel="voxel count", xlabel="normalized intensity", ylim=(10, 500000), linewidth=1, legend=:topleft, yscale=yscale, size=(500, 300))
plot!([thresholds[1], thresholds[1]], [1, 10000_000], label="lower threshold", linewidth=3)
plot!([thresholds[2], thresholds[2]], [1, 10000_000], label="upper threshold", linewidth=3)
#plot!([chosen_threshold, chosen_threshold], [1, 30000000], label="chosen threshold", linewidth=3)
end
# ╔═╡ 4e3274ed-ad71-4568-a930-c50abe7c6bd5
plot_histogram(Array(target), Array(fwd(res.minimizer)), (0.65, 0.75))
# ╔═╡ 03223eb4-3c15-4899-93bf-5440fb6b5c7a
plot_histogram(Array(target), Array(printed), (0.65, 0.75))
# ╔═╡ Cell order:
# ╠═179d107e-b3be-11ee-0a6c-49f1bf2a10fd
# ╟─eb4b7177-4f93-4728-8c91-4ff6a86f40cf
# ╠═e3a3dc83-6536-445f-b56b-d8ad62cd9a85
# ╠═f3277f00-4162-43b2-98e0-53bcb508df1e
# ╠═42f31ceb-0f5d-46cc-b868-0314b41a5407
# ╟─78746412-00f2-40ad-9d24-e884d8dd7e77
# ╟─980e9656-202d-4a47-837f-b2197fe5d3b4
# ╠═f0d86933-5eab-4d86-8909-c51ab9ca99d9
# ╠═36eab63f-5643-4b43-8e12-0f7d971c1040
# ╠═fc049fbc-4ae7-4256-b39f-27fdff98d45a
# ╠═a6bdc90f-3c10-4524-9892-6e976d556a95
# ╟─066664f7-9ae5-4a4d-b541-663067cea361
# ╠═5ef75bb5-5ab0-4e98-9026-a8e2c5b7833d
# ╠═5aa03590-c552-410e-93aa-1e47dabd2f9e
# ╠═9d582f0d-7da4-41db-9369-fef7a6dbe590
# ╟─6b91d2bd-1170-4ce5-bb20-10d69f8cd6d7
# ╠═8f212882-5a4f-4915-a01a-9c7f342ea1ef
# ╠═07ff2197-836d-4e9e-b716-4e0eacd89e6e
# ╠═6deeaf05-1008-4f1e-bb33-2b74196c9d80
# ╠═b94f82e8-1c24-46a0-b3e9-f98a0e3e1c12
# ╠═66a09fd3-545b-4456-be74-6593d2d15a28
# ╠═6721b65f-48ed-4dc7-99dd-c460bb124c1e
# ╠═76b4c0de-af7a-456e-8d49-04f10a0a17fb
# ╠═af83cb81-49b8-41f4-a24e-c1e33e76d925
# ╟─4cdc5592-3b92-4e98-8718-d77dd939ebf1
# ╠═dc22dead-1b83-4787-a147-9811136af085
# ╠═494c5b20-0af2-4903-959c-812459877c31
# ╠═e10549a7-0f31-48fa-8d38-820907d8554e
# ╠═e83be634-dde4-4ed4-9226-2ba895a84d15
# ╠═85b6e54c-3256-4ba8-9889-56be82de0203
# ╟─5d55464f-c7d9-4e45-a1e7-907c2137be95
# ╠═834cd0f6-9be0-4589-82c0-ca0021911dd5
# ╠═697d4c7d-697b-452c-8a6a-25f2a1089529
# ╠═624858de-e7da-4e3c-91b5-6ab7b5e645d5
# ╠═12e7c6b7-fcf3-44e4-9d80-785704d93cb5
# ╠═942e1cf7-a47b-43af-a460-1ea896a2c9b8
# ╟─d80f6c47-1c55-4a7e-9374-104806873a9a
# ╠═be45bfe1-e91e-45eb-8902-62e8df0d8ac7
# ╠═4e3274ed-ad71-4568-a930-c50abe7c6bd5
# ╠═e14a65c3-ece0-4d1e-96a7-9a9ab00001ad
# ╠═2ca8698e-3ff8-445c-a5e3-3a20d2e2afc3
# ╠═03223eb4-3c15-4899-93bf-5440fb6b5c7a
# ╠═ec528b63-95eb-430b-b683-04e207dd61f0
# ╠═43e4b941-b838-483d-9a8f-8897720f8b90
# ╠═45884ebb-a02d-446a-a7ca-05b8b4abc244
# ╠═7c998816-b4de-4dbf-977b-bdab9355be79
# ╠═28d8c378-b818-4d61-b9d4-a06ad7cc21e3
# ╠═beb61bb1-3f7c-4301-92ca-5418bc0445cf
# ╠═61048e10-8e23-41a6-8c49-0b1a8c0adc24
# ╠═88895d8d-432e-4337-a51c-12ac4e4a4325
# ╠═21c0b369-a4b2-43ea-87b8-702474d88584
# ╠═b53c8016-177f-4b1b-b07d-48cbd2b2fc65